@cursor/july 0.1.1 → 0.1.3

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (298) hide show
  1. package/AGENTS.md +40 -7
  2. package/README.md +33 -24
  3. package/dist/bin/agent-serve.d.ts +3 -1
  4. package/dist/bin/agent-serve.d.ts.map +1 -1
  5. package/dist/bin/agent-serve.js +466 -140
  6. package/dist/channels/github/api.d.ts.map +1 -1
  7. package/dist/channels/github/api.js +31 -14
  8. package/dist/channels/github/cursor-account.d.ts +43 -0
  9. package/dist/channels/github/cursor-account.d.ts.map +1 -0
  10. package/dist/channels/github/cursor-account.js +95 -0
  11. package/dist/channels/github/github-channel.d.ts.map +1 -1
  12. package/dist/channels/github/github-channel.js +46 -9
  13. package/dist/channels/github/index.d.ts +2 -2
  14. package/dist/channels/github/index.js +2 -2
  15. package/dist/channels/github/types.d.ts +17 -0
  16. package/dist/channels/github/types.d.ts.map +1 -1
  17. package/dist/channels/slack/slack-channel.d.ts +8 -2
  18. package/dist/channels/slack/slack-channel.d.ts.map +1 -1
  19. package/dist/channels/slack/slack-channel.js +8 -0
  20. package/dist/channels/slack/types.d.ts +24 -3
  21. package/dist/channels/slack/types.d.ts.map +1 -1
  22. package/dist/channels/slack/types.js +15 -1
  23. package/dist/docs/404.html +2 -2
  24. package/dist/docs/ab.html +8 -8
  25. package/dist/docs/assets/{ab.md.COdXkces.js → ab.md.BMCZ6Hd7.js} +3 -3
  26. package/dist/docs/assets/{ab.md.COdXkces.lean.js → ab.md.BMCZ6Hd7.lean.js} +1 -1
  27. package/dist/docs/assets/{app.DqfFEmJd.js → app.BR0RbIdx.js} +1 -1
  28. package/dist/docs/assets/chunks/@localSearchIndexroot.DtIOQzGj.js +1 -0
  29. package/dist/docs/assets/chunks/{VPLocalSearchBox.BaLEdS15.js → VPLocalSearchBox.qjsVTneI.js} +1 -1
  30. package/dist/docs/assets/chunks/{theme.CZRvu_0q.js → theme.2Kg8jIp_.js} +2 -2
  31. package/dist/docs/assets/{deployment.md.Dx1TYNk5.js → deployment.md.DTKwE15Z.js} +3 -3
  32. package/dist/docs/assets/{deployment.md.Dx1TYNk5.lean.js → deployment.md.DTKwE15Z.lean.js} +1 -1
  33. package/dist/docs/assets/{evals.md.DPZ_MAnI.js → evals.md.DAgEc_hL.js} +3 -3
  34. package/dist/docs/assets/{guides_agent-to-agent.md.CrtrsySy.js → guides_agent-to-agent.md.Bpzgq2Pq.js} +1 -1
  35. package/dist/docs/assets/{guides_cloud-runtime.md.CYlNMTNp.js → guides_cloud-runtime.md.gVzabdQL.js} +1 -1
  36. package/dist/docs/assets/{guides_github.md.DwbKhCeS.js → guides_github.md.DOOCpqsW.js} +11 -4
  37. package/dist/docs/assets/{guides_github.md.DwbKhCeS.lean.js → guides_github.md.DOOCpqsW.lean.js} +1 -1
  38. package/dist/docs/assets/{guides_human-in-the-loop.md.Dvuctx7s.js → guides_human-in-the-loop.md.DlUqsp1S.js} +2 -2
  39. package/dist/docs/assets/{guides_slack.md.bv41fHfW.js → guides_slack.md.CCwqHvSV.js} +4 -4
  40. package/dist/docs/assets/{guides_slack.md.bv41fHfW.lean.js → guides_slack.md.CCwqHvSV.lean.js} +1 -1
  41. package/dist/docs/assets/{guides_webhooks.md.hFTik3lf.js → guides_webhooks.md.B1EswtUu.js} +2 -2
  42. package/dist/docs/assets/index.md.m81y7TY7.js +20 -0
  43. package/dist/docs/assets/{index.md.BPKcj5AI.lean.js → index.md.m81y7TY7.lean.js} +1 -1
  44. package/dist/docs/assets/quickstart.md.BU6Iwi_9.js +204 -0
  45. package/dist/docs/assets/quickstart.md.BU6Iwi_9.lean.js +1 -0
  46. package/dist/docs/assets/{reference_agent-config.md.Bpd7HQwf.js → reference_agent-config.md.DrW2JUM8.js} +4 -4
  47. package/dist/docs/assets/{reference_agent-config.md.Bpd7HQwf.lean.js → reference_agent-config.md.DrW2JUM8.lean.js} +1 -1
  48. package/dist/docs/assets/{reference_channels.md.D7JTR03W.js → reference_channels.md.DdmiKgqf.js} +4 -4
  49. package/dist/docs/assets/{reference_channels.md.D7JTR03W.lean.js → reference_channels.md.DdmiKgqf.lean.js} +1 -1
  50. package/dist/docs/assets/{reference_cli.md.DA730zCu.js → reference_cli.md.Bv6pOxcF.js} +11 -6
  51. package/dist/docs/assets/{reference_cli.md.DA730zCu.lean.js → reference_cli.md.Bv6pOxcF.lean.js} +1 -1
  52. package/dist/docs/assets/{reference_connections.md.C3vNH_DE.js → reference_connections.md.zaEYCLHT.js} +1 -1
  53. package/dist/docs/assets/{reference_hooks.md.BCEc3MyM.js → reference_hooks.md.DyLVfE1O.js} +1 -1
  54. package/dist/docs/assets/{reference_hooks.md.BCEc3MyM.lean.js → reference_hooks.md.DyLVfE1O.lean.js} +1 -1
  55. package/dist/docs/assets/{reference_http-api.md.DBAahtdz.js → reference_http-api.md.Dx_nmDG6.js} +1 -1
  56. package/dist/docs/assets/{reference_instructions.md.BC05LEQ8.js → reference_instructions.md.CgoV-YEb.js} +9 -7
  57. package/dist/docs/assets/{reference_instructions.md.BC05LEQ8.lean.js → reference_instructions.md.CgoV-YEb.lean.js} +1 -1
  58. package/dist/docs/assets/{reference_schedules.md.D7qijxLk.js → reference_schedules.md.w_F2mXB6.js} +2 -2
  59. package/dist/docs/assets/{reference_skills.md.VQnlBT3Q.js → reference_skills.md.B_jHN7JL.js} +3 -3
  60. package/dist/docs/assets/{reference_skills.md.VQnlBT3Q.lean.js → reference_skills.md.B_jHN7JL.lean.js} +1 -1
  61. package/dist/docs/assets/{reference_subagents.md.CIRAVcPK.js → reference_subagents.md.zWAMNfi1.js} +1 -1
  62. package/dist/docs/assets/{reference_tools.md.DF5kwlt0.js → reference_tools.md.CqgJroI0.js} +2 -2
  63. package/dist/docs/assets/scaffolding-agents.md.C3pTrmoE.js +1 -0
  64. package/dist/docs/assets/scaffolding-agents.md.C3pTrmoE.lean.js +1 -0
  65. package/dist/docs/assets/storage.md.CVnInNiN.js +17 -0
  66. package/dist/docs/assets/storage.md.CVnInNiN.lean.js +1 -0
  67. package/dist/docs/building-with-agents.html +4 -4
  68. package/dist/docs/concepts.html +4 -4
  69. package/dist/docs/deployment.html +7 -7
  70. package/dist/docs/evals.html +7 -7
  71. package/dist/docs/guides/agent-to-agent.html +6 -6
  72. package/dist/docs/guides/cloud-runtime.html +5 -5
  73. package/dist/docs/guides/github.html +14 -7
  74. package/dist/docs/guides/human-in-the-loop.html +6 -6
  75. package/dist/docs/guides/slack.html +8 -8
  76. package/dist/docs/guides/webhooks.html +6 -6
  77. package/dist/docs/hashmap.json +1 -1
  78. package/dist/docs/hillclimbing.html +5 -5
  79. package/dist/docs/index.html +7 -7
  80. package/dist/docs/quickstart.html +195 -26
  81. package/dist/docs/reference/agent-config.html +7 -7
  82. package/dist/docs/reference/channels.html +8 -8
  83. package/dist/docs/reference/cli.html +14 -9
  84. package/dist/docs/reference/connections.html +5 -5
  85. package/dist/docs/reference/hooks.html +5 -5
  86. package/dist/docs/reference/http-api.html +6 -6
  87. package/dist/docs/reference/instructions.html +13 -11
  88. package/dist/docs/reference/playground.html +4 -4
  89. package/dist/docs/reference/project-layout.html +4 -4
  90. package/dist/docs/reference/schedules.html +7 -7
  91. package/dist/docs/reference/sessions.html +4 -4
  92. package/dist/docs/reference/skills.html +6 -6
  93. package/dist/docs/reference/subagents.html +6 -6
  94. package/dist/docs/reference/tools.html +7 -7
  95. package/dist/docs/scaffolding-agents.html +5 -5
  96. package/dist/docs/storage.html +41 -0
  97. package/dist/docs/troubleshooting.html +4 -4
  98. package/dist/index.d.ts +2 -0
  99. package/dist/index.d.ts.map +1 -1
  100. package/dist/index.js +1 -0
  101. package/dist/internal/chat-client.d.ts +29 -3
  102. package/dist/internal/chat-client.d.ts.map +1 -1
  103. package/dist/internal/chat-client.js +180 -27
  104. package/dist/internal/cli-ax.d.ts +75 -2
  105. package/dist/internal/cli-ax.d.ts.map +1 -1
  106. package/dist/internal/cli-ax.js +600 -52
  107. package/dist/internal/cli-cursor.d.ts.map +1 -1
  108. package/dist/internal/cli-cursor.js +8 -36
  109. package/dist/internal/cli-deploy.d.ts +108 -0
  110. package/dist/internal/cli-deploy.d.ts.map +1 -0
  111. package/dist/internal/cli-deploy.js +1009 -0
  112. package/dist/internal/cursor/backend-client.d.ts +10 -1
  113. package/dist/internal/cursor/backend-client.d.ts.map +1 -1
  114. package/dist/internal/cursor/backend-client.js +79 -1
  115. package/dist/internal/cursor/credentials.d.ts +9 -0
  116. package/dist/internal/cursor/credentials.d.ts.map +1 -1
  117. package/dist/internal/cursor/credentials.js +12 -0
  118. package/dist/internal/cursor/github-credentials.d.ts +44 -0
  119. package/dist/internal/cursor/github-credentials.d.ts.map +1 -0
  120. package/dist/internal/cursor/github-credentials.js +195 -0
  121. package/dist/internal/deploy-client.d.ts +184 -0
  122. package/dist/internal/deploy-client.d.ts.map +1 -0
  123. package/dist/internal/deploy-client.js +395 -0
  124. package/dist/internal/deploy-source.d.ts +35 -0
  125. package/dist/internal/deploy-source.d.ts.map +1 -0
  126. package/dist/internal/deploy-source.js +118 -0
  127. package/dist/internal/discovery.d.ts +11 -0
  128. package/dist/internal/discovery.d.ts.map +1 -1
  129. package/dist/internal/discovery.js +116 -19
  130. package/dist/internal/distribution.d.ts.map +1 -1
  131. package/dist/internal/distribution.js +1 -0
  132. package/dist/internal/eval-run-store.d.ts +22 -3
  133. package/dist/internal/eval-run-store.d.ts.map +1 -1
  134. package/dist/internal/eval-run-store.js +39 -20
  135. package/dist/internal/eval-runner.d.ts +2 -0
  136. package/dist/internal/eval-runner.d.ts.map +1 -1
  137. package/dist/internal/eval-runner.js +2 -0
  138. package/dist/internal/event-mapper.d.ts +54 -1
  139. package/dist/internal/event-mapper.d.ts.map +1 -1
  140. package/dist/internal/event-mapper.js +151 -41
  141. package/dist/internal/handleAgentServeTrigger.d.ts.map +1 -1
  142. package/dist/internal/handleAgentServeTrigger.js +10 -12
  143. package/dist/internal/host-platforms.d.ts +7 -2
  144. package/dist/internal/host-platforms.d.ts.map +1 -1
  145. package/dist/internal/host-platforms.js +15 -10
  146. package/dist/internal/hosting.d.ts +37 -0
  147. package/dist/internal/hosting.d.ts.map +1 -0
  148. package/dist/internal/hosting.js +67 -0
  149. package/dist/internal/init-project.d.ts +88 -1
  150. package/dist/internal/init-project.d.ts.map +1 -1
  151. package/dist/internal/init-project.js +221 -31
  152. package/dist/internal/install-cursor-skills.d.ts +64 -0
  153. package/dist/internal/install-cursor-skills.d.ts.map +1 -0
  154. package/dist/internal/install-cursor-skills.js +274 -0
  155. package/dist/internal/logs-client.d.ts +60 -0
  156. package/dist/internal/logs-client.d.ts.map +1 -0
  157. package/dist/internal/logs-client.js +311 -0
  158. package/dist/internal/open-browser.d.ts +5 -0
  159. package/dist/internal/open-browser.d.ts.map +1 -0
  160. package/dist/internal/open-browser.js +34 -0
  161. package/dist/internal/playground/toolchain.d.ts.map +1 -1
  162. package/dist/internal/playground/toolchain.js +16 -11
  163. package/dist/internal/playground-proxy.d.ts +69 -0
  164. package/dist/internal/playground-proxy.d.ts.map +1 -0
  165. package/dist/internal/playground-proxy.js +468 -0
  166. package/dist/internal/prompt-context.d.ts +35 -0
  167. package/dist/internal/prompt-context.d.ts.map +1 -0
  168. package/dist/internal/prompt-context.js +71 -0
  169. package/dist/internal/reminder-runner.d.ts +7 -0
  170. package/dist/internal/reminder-runner.d.ts.map +1 -1
  171. package/dist/internal/reminder-runner.js +50 -6
  172. package/dist/internal/reminder-store.d.ts +2 -0
  173. package/dist/internal/reminder-store.d.ts.map +1 -1
  174. package/dist/internal/reminder-store.js +18 -0
  175. package/dist/internal/request-headers.d.ts +11 -0
  176. package/dist/internal/request-headers.d.ts.map +1 -0
  177. package/dist/internal/request-headers.js +14 -0
  178. package/dist/internal/resolve-prod-target.d.ts +42 -0
  179. package/dist/internal/resolve-prod-target.d.ts.map +1 -0
  180. package/dist/internal/resolve-prod-target.js +111 -0
  181. package/dist/internal/run-client.d.ts +2 -2
  182. package/dist/internal/run-client.d.ts.map +1 -1
  183. package/dist/internal/run-client.js +38 -23
  184. package/dist/internal/server.d.ts.map +1 -1
  185. package/dist/internal/server.js +182 -45
  186. package/dist/internal/session-engine.d.ts +50 -1
  187. package/dist/internal/session-engine.d.ts.map +1 -1
  188. package/dist/internal/session-engine.js +267 -23
  189. package/dist/internal/sessions-client.d.ts +21 -0
  190. package/dist/internal/sessions-client.d.ts.map +1 -0
  191. package/dist/internal/sessions-client.js +168 -0
  192. package/dist/internal/storage-coordinator.d.ts +139 -0
  193. package/dist/internal/storage-coordinator.d.ts.map +1 -0
  194. package/dist/internal/storage-coordinator.js +499 -0
  195. package/dist/internal/stream-progress.d.ts.map +1 -1
  196. package/dist/internal/stream-progress.js +31 -3
  197. package/dist/internal/terminal-style.d.ts +22 -0
  198. package/dist/internal/terminal-style.d.ts.map +1 -0
  199. package/dist/internal/terminal-style.js +49 -0
  200. package/dist/internal/trajectory.d.ts +10 -5
  201. package/dist/internal/trajectory.d.ts.map +1 -1
  202. package/dist/internal/trajectory.js +82 -10
  203. package/dist/internal/workspace.d.ts +11 -0
  204. package/dist/internal/workspace.d.ts.map +1 -1
  205. package/dist/internal/workspace.js +38 -4
  206. package/dist/playground/assets/cursor-icons-outline-BxTT_FVJ.woff2 +0 -0
  207. package/dist/playground/assets/index-D-vo2lV_.css +1 -0
  208. package/dist/playground/assets/index-x60b9q2j.js +312 -0
  209. package/dist/playground/index.html +2 -2
  210. package/dist/storage.d.ts +204 -0
  211. package/dist/storage.d.ts.map +1 -0
  212. package/dist/storage.js +153 -0
  213. package/dist/types.d.ts +70 -4
  214. package/dist/types.d.ts.map +1 -1
  215. package/docs/README.md +3 -2
  216. package/docs/guides/github.md +43 -8
  217. package/docs/guides/slack.md +1 -1
  218. package/docs/quickstart.md +350 -57
  219. package/docs/reference/cli.md +51 -3
  220. package/docs/reference/instructions.md +8 -6
  221. package/docs/scaffolding-agents.md +1 -1
  222. package/docs/storage.md +98 -0
  223. package/package.json +10 -1
  224. package/skills/ab/SKILL.md +8 -8
  225. package/skills/create-agent/SKILL.md +28 -22
  226. package/skills/debug/SKILL.md +11 -11
  227. package/skills/evals/SKILL.md +16 -16
  228. package/skills/framework-map/SKILL.md +6 -6
  229. package/skills/github/SKILL.md +22 -14
  230. package/skills/hillclimb/SKILL.md +10 -10
  231. package/skills/setup-slack/SKILL.md +13 -13
  232. package/src/bin/agent-serve.ts +552 -179
  233. package/src/channels/github/api.ts +42 -23
  234. package/src/channels/github/cursor-account.ts +165 -0
  235. package/src/channels/github/github-channel.ts +66 -6
  236. package/src/channels/github/index.ts +2 -2
  237. package/src/channels/github/types.ts +19 -0
  238. package/src/channels/slack/slack-channel.ts +17 -3
  239. package/src/channels/slack/types.ts +44 -3
  240. package/src/index.ts +13 -0
  241. package/src/internal/chat-client.ts +252 -37
  242. package/src/internal/cli-ax.ts +722 -72
  243. package/src/internal/cli-cursor.ts +14 -42
  244. package/src/internal/cli-deploy.ts +1319 -0
  245. package/src/internal/cursor/backend-client.ts +103 -1
  246. package/src/internal/cursor/credentials.ts +14 -0
  247. package/src/internal/cursor/github-credentials.ts +248 -0
  248. package/src/internal/deploy-client.ts +632 -0
  249. package/src/internal/deploy-source.ts +133 -0
  250. package/src/internal/discovery.ts +141 -17
  251. package/src/internal/distribution.ts +1 -0
  252. package/src/internal/eval-run-store.ts +50 -20
  253. package/src/internal/eval-runner.ts +5 -0
  254. package/src/internal/event-mapper.ts +222 -42
  255. package/src/internal/handleAgentServeTrigger.ts +10 -12
  256. package/src/internal/host-platforms.ts +28 -11
  257. package/src/internal/hosting.ts +77 -0
  258. package/src/internal/init-project.ts +333 -51
  259. package/src/internal/install-cursor-skills.ts +327 -0
  260. package/src/internal/logs-client.ts +442 -0
  261. package/src/internal/open-browser.ts +41 -0
  262. package/src/internal/playground/toolchain.ts +15 -13
  263. package/src/internal/playground-proxy.ts +576 -0
  264. package/src/internal/prompt-context.ts +95 -0
  265. package/src/internal/reminder-runner.ts +50 -6
  266. package/src/internal/reminder-store.ts +21 -0
  267. package/src/internal/request-headers.ts +23 -0
  268. package/src/internal/resolve-prod-target.ts +150 -0
  269. package/src/internal/run-client.ts +39 -36
  270. package/src/internal/server.ts +223 -28
  271. package/src/internal/session-engine.ts +307 -10
  272. package/src/internal/sessions-client.ts +182 -0
  273. package/src/internal/storage-coordinator.ts +615 -0
  274. package/src/internal/stream-progress.ts +36 -2
  275. package/src/internal/terminal-style.ts +74 -0
  276. package/src/internal/trajectory.ts +91 -13
  277. package/src/internal/workspace.ts +42 -4
  278. package/src/storage.ts +325 -0
  279. package/src/types.ts +82 -8
  280. package/dist/docs/assets/chunks/@localSearchIndexroot.CcVk1uKq.js +0 -1
  281. package/dist/docs/assets/index.md.BPKcj5AI.js +0 -20
  282. package/dist/docs/assets/quickstart.md.tVPiGK_L.js +0 -35
  283. package/dist/docs/assets/quickstart.md.tVPiGK_L.lean.js +0 -1
  284. package/dist/docs/assets/scaffolding-agents.md.CyYfWGdc.js +0 -1
  285. package/dist/docs/assets/scaffolding-agents.md.CyYfWGdc.lean.js +0 -1
  286. package/dist/playground/assets/cursor-icons-outline-oY2V_mvK.woff2 +0 -0
  287. package/dist/playground/assets/index-1K-hG-7p.css +0 -1
  288. package/dist/playground/assets/index-FlWjhg3x.js +0 -79
  289. /package/dist/docs/assets/{evals.md.DPZ_MAnI.lean.js → evals.md.DAgEc_hL.lean.js} +0 -0
  290. /package/dist/docs/assets/{guides_agent-to-agent.md.CrtrsySy.lean.js → guides_agent-to-agent.md.Bpzgq2Pq.lean.js} +0 -0
  291. /package/dist/docs/assets/{guides_cloud-runtime.md.CYlNMTNp.lean.js → guides_cloud-runtime.md.gVzabdQL.lean.js} +0 -0
  292. /package/dist/docs/assets/{guides_human-in-the-loop.md.Dvuctx7s.lean.js → guides_human-in-the-loop.md.DlUqsp1S.lean.js} +0 -0
  293. /package/dist/docs/assets/{guides_webhooks.md.hFTik3lf.lean.js → guides_webhooks.md.B1EswtUu.lean.js} +0 -0
  294. /package/dist/docs/assets/{reference_connections.md.C3vNH_DE.lean.js → reference_connections.md.zaEYCLHT.lean.js} +0 -0
  295. /package/dist/docs/assets/{reference_http-api.md.DBAahtdz.lean.js → reference_http-api.md.Dx_nmDG6.lean.js} +0 -0
  296. /package/dist/docs/assets/{reference_schedules.md.D7qijxLk.lean.js → reference_schedules.md.w_F2mXB6.lean.js} +0 -0
  297. /package/dist/docs/assets/{reference_subagents.md.CIRAVcPK.lean.js → reference_subagents.md.zWAMNfi1.lean.js} +0 -0
  298. /package/dist/docs/assets/{reference_tools.md.DF5kwlt0.lean.js → reference_tools.md.CqgJroI0.lean.js} +0 -0
@@ -9,17 +9,17 @@
9
9
  <link rel="preload stylesheet" href="/docs/assets/style.Bj-IXa5W.css" as="style">
10
10
  <link rel="preload stylesheet" href="/docs/vp-icons.css" as="style">
11
11
 
12
- <script type="module" src="/docs/assets/app.DqfFEmJd.js"></script>
12
+ <script type="module" src="/docs/assets/app.BR0RbIdx.js"></script>
13
13
  <link rel="preload" href="/docs/assets/inter-roman-latin.Di8DUHzh.woff2" as="font" type="font/woff2" crossorigin="">
14
- <link rel="modulepreload" href="/docs/assets/chunks/theme.CZRvu_0q.js">
14
+ <link rel="modulepreload" href="/docs/assets/chunks/theme.2Kg8jIp_.js">
15
15
  <link rel="modulepreload" href="/docs/assets/chunks/framework.CAZyNGu9.js">
16
- <link rel="modulepreload" href="/docs/assets/evals.md.DPZ_MAnI.lean.js">
16
+ <link rel="modulepreload" href="/docs/assets/evals.md.DAgEc_hL.lean.js">
17
17
  <script id="check-dark-mode">(()=>{const e=localStorage.getItem("vitepress-theme-appearance")||"auto",a=window.matchMedia("(prefers-color-scheme: dark)").matches;(!e||e==="auto"?a:e==="dark")&&document.documentElement.classList.add("dark")})();</script>
18
18
  <script id="check-mac-os">document.documentElement.classList.toggle("mac",/Mac|iPhone|iPod|iPad/i.test(navigator.platform));</script>
19
19
  </head>
20
20
  <body>
21
- <div id="app"><div class="Layout" data-v-05085060><!--[--><!--]--><!--[--><span tabindex="-1" data-v-30dc752a></span><a href="#VPContent" class="VPSkipLink visually-hidden" data-v-30dc752a>Skip to content</a><!--]--><!----><header class="VPNav" data-v-05085060 data-v-cbe59881><div class="VPNavBar" data-v-cbe59881 data-v-dd8a1864><div class="wrapper" data-v-dd8a1864><div class="container" data-v-dd8a1864><div class="title" data-v-dd8a1864><div class="VPNavBarTitle has-sidebar" data-v-dd8a1864 data-v-f9346b2b><a class="title" href="/docs/" data-v-f9346b2b><!--[--><!--]--><!----><span data-v-f9346b2b>agentkit</span><!--[--><!--]--></a></div></div><div class="content" data-v-dd8a1864><div class="content-body" data-v-dd8a1864><!--[--><!--]--><div class="VPNavBarSearch search" data-v-dd8a1864><!--[--><!----><div id="local-search"><button type="button" class="DocSearch DocSearch-Button" aria-label="Search"><span class="DocSearch-Button-Container"><span class="vp-icon DocSearch-Search-Icon"></span><span class="DocSearch-Button-Placeholder">Search</span></span><span class="DocSearch-Button-Keys"><kbd class="DocSearch-Button-Key"></kbd><kbd class="DocSearch-Button-Key">K</kbd></span></button></div><!--]--></div><nav aria-labelledby="main-nav-aria-label" class="VPNavBarMenu menu" data-v-dd8a1864 data-v-7ed6e5ac><span id="main-nav-aria-label" class="visually-hidden" data-v-7ed6e5ac> Main Navigation </span><!--[--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/quickstart.html" tabindex="0" data-v-7ed6e5ac data-v-65720067><!--[--><span data-v-65720067>Quickstart</span><!--]--></a><!--]--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/guides/webhooks.html" tabindex="0" data-v-7ed6e5ac data-v-65720067><!--[--><span data-v-65720067>Guides</span><!--]--></a><!--]--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/reference/project-layout.html" tabindex="0" data-v-7ed6e5ac data-v-65720067><!--[--><span data-v-65720067>Reference</span><!--]--></a><!--]--><!--]--></nav><!----><div class="VPNavBarAppearance appearance" data-v-dd8a1864 data-v-2b2b395e><button class="VPSwitch VPSwitchAppearance" type="button" role="switch" title aria-checked="false" data-v-2b2b395e data-v-182deabf data-v-2562b538><span class="check" data-v-2562b538><span class="icon" data-v-2562b538><!--[--><span class="vpi-sun sun" data-v-182deabf></span><span class="vpi-moon moon" data-v-182deabf></span><!--]--></span></span></button></div><!----><div class="VPFlyout VPNavBarExtra extra" data-v-dd8a1864 data-v-f3b584b7 data-v-82133de6><button type="button" class="button" aria-haspopup="true" aria-expanded="false" aria-label="extra navigation" data-v-82133de6><span class="vpi-more-horizontal icon" data-v-82133de6></span></button><div class="menu" data-v-82133de6><div class="VPMenu" data-v-82133de6 data-v-759c9e19><!----><!--[--><!--[--><!----><div class="group" data-v-f3b584b7><div class="item appearance" data-v-f3b584b7><p class="label" data-v-f3b584b7>Appearance</p><div class="appearance-action" data-v-f3b584b7><button class="VPSwitch VPSwitchAppearance" type="button" role="switch" title aria-checked="false" data-v-f3b584b7 data-v-182deabf data-v-2562b538><span class="check" data-v-2562b538><span class="icon" data-v-2562b538><!--[--><span class="vpi-sun sun" data-v-182deabf></span><span class="vpi-moon moon" data-v-182deabf></span><!--]--></span></span></button></div></div></div><!----><!--]--><!--]--></div></div></div><!--[--><!--]--><button type="button" class="VPNavBarHamburger hamburger" aria-label="mobile navigation" aria-expanded="false" aria-controls="VPNavScreen" data-v-dd8a1864 data-v-ab07e844><span class="container" data-v-ab07e844><span class="top" data-v-ab07e844></span><span class="middle" data-v-ab07e844></span><span class="bottom" data-v-ab07e844></span></span></button></div></div></div></div><div class="divider" data-v-dd8a1864><div class="divider-line" data-v-dd8a1864></div></div></div><!----></header><div class="VPLocalNav has-sidebar empty" data-v-05085060 data-v-826b8761><div class="container" data-v-826b8761><button class="menu" aria-expanded="false" aria-controls="VPSidebarNav" data-v-826b8761><span class="vpi-align-left menu-icon" data-v-826b8761></span><span class="menu-text" data-v-826b8761>Menu</span></button><div class="VPLocalNavOutlineDropdown" style="--vp-vh:0px;" data-v-826b8761 data-v-01af9e73><button data-v-01af9e73>Return to top</button><!----></div></div></div><aside class="VPSidebar" data-v-05085060 data-v-5cd5b4d8><div class="curtain" data-v-5cd5b4d8></div><nav class="nav" id="VPSidebarNav" aria-labelledby="sidebar-aria-label" tabindex="-1" data-v-5cd5b4d8><span class="visually-hidden" id="sidebar-aria-label" data-v-5cd5b4d8> Sidebar Navigation </span><!--[--><!--]--><!--[--><div class="no-transition group" data-v-6b69f472><section class="VPSidebarItem level-0" data-v-6b69f472 data-v-b6893b1b><div class="item" role="button" tabindex="0" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><h2 class="text" data-v-b6893b1b>Getting started</h2><!----></div><div class="items" data-v-b6893b1b><!--[--><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Overview</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/quickstart.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Quickstart</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/scaffolding-agents.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Scaffold an agent with Cursor</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/concepts.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Concepts</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-6b69f472><section class="VPSidebarItem level-0 has-active" data-v-6b69f472 data-v-b6893b1b><div class="item" role="button" tabindex="0" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><h2 class="text" data-v-b6893b1b>Improving agents</h2><!----></div><div class="items" data-v-b6893b1b><!--[--><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/building-with-agents.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Building agents with agents</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/evals.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Evals</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/ab.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Live A/B metrics</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/hillclimbing.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Hillclimbing</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-6b69f472><section class="VPSidebarItem level-0" data-v-6b69f472 data-v-b6893b1b><div class="item" role="button" tabindex="0" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><h2 class="text" data-v-b6893b1b>Guides</h2><!----></div><div class="items" data-v-b6893b1b><!--[--><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/guides/webhooks.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Webhooks & custom channels</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/guides/github.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>GitHub</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/guides/slack.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Slack</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/guides/human-in-the-loop.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Human-in-the-loop approvals</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/guides/agent-to-agent.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Agent-to-agent</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/guides/cloud-runtime.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Cloud runtime</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-6b69f472><section class="VPSidebarItem level-0" data-v-6b69f472 data-v-b6893b1b><div class="item" role="button" tabindex="0" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><h2 class="text" data-v-b6893b1b>Operating</h2><!----></div><div class="items" data-v-b6893b1b><!--[--><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/deployment.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Deployment</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/troubleshooting.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Fix common problems</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-6b69f472><section class="VPSidebarItem level-0 collapsible" data-v-6b69f472 data-v-b6893b1b><div class="item" role="button" tabindex="0" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><h2 class="text" data-v-b6893b1b>Reference</h2><div class="caret" role="button" aria-label="toggle section" tabindex="0" data-v-b6893b1b><span class="vpi-chevron-right caret-icon" data-v-b6893b1b></span></div></div><div class="items" data-v-b6893b1b><!--[--><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/project-layout.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Project layout</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/agent-config.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Agent config</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/instructions.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Instructions</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/tools.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Tools</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/skills.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Skills</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/connections.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>MCP Connections</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/subagents.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Subagents</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/channels.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Channels</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/schedules.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Schedules & reminders</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/hooks.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Hooks</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/sessions.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Sessions & streaming</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/playground.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Playground</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/cli.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>CLI</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/http-api.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>HTTP API</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><!--]--><!--[--><!--]--></nav></aside><div class="VPContent has-sidebar" id="VPContent" data-v-05085060 data-v-8eedf2ac><div class="VPDoc has-sidebar has-aside" data-v-8eedf2ac data-v-53b9fda6><!--[--><!--]--><div class="container" data-v-53b9fda6><div class="aside" data-v-53b9fda6><div class="aside-curtain" data-v-53b9fda6></div><div class="aside-container" data-v-53b9fda6><div class="aside-content" data-v-53b9fda6><div class="VPDocAside" data-v-53b9fda6 data-v-e448ecab><!--[--><!--]--><!--[--><!--]--><nav aria-labelledby="doc-outline-aria-label" class="VPDocAsideOutline" data-v-e448ecab data-v-b0b42374><div class="content" data-v-b0b42374><div class="outline-marker" data-v-b0b42374></div><div aria-level="2" class="outline-title" id="doc-outline-aria-label" role="heading" data-v-b0b42374>On this page</div><ul class="VPDocOutlineItem root" data-v-b0b42374 data-v-ae03bd52><!--[--><!--]--></ul></div></nav><!--[--><!--]--><div class="spacer" data-v-e448ecab></div><!--[--><!--]--><!----><!--[--><!--]--><!--[--><!--]--></div></div></div></div><div class="content" data-v-53b9fda6><div class="content-container" data-v-53b9fda6><!--[--><!--]--><main class="main" data-v-53b9fda6><div style="position:relative;" class="vp-doc _docs_evals" data-v-53b9fda6><div><h1 id="evals" tabindex="-1">Evals <a class="header-anchor" href="#evals" aria-label="Permalink to &quot;Evals&quot;">​</a></h1><p>An eval is a repeatable check that runs your agent against a fixed input and gates the recorded trajectory: the run completed, the right tool ran, the reply has the right shape. Evals are how you know a prompt tweak helped, a refactor didn&#39;t regress the agent, and last month&#39;s fix is still holding.</p><p>Evals exercise the same surface your users hit. The runner starts (or targets) a real agent server, drives sessions over the public API, and grades what comes back. A passing eval means the agent started, accepted a message, and did what you asserted.</p><div class="note custom-block github-alert"><p class="custom-block-title">NOTE</p><p>Import paths here use <code>@anysphere/agentkit/evals</code>. On projects still using <code>@anysphere/agent-serve</code>, swap the import and run <code>agent-serve eval</code>. See <a href="./README.html#run-the-cli">Run the CLI</a> for the full rename table.</p></div><h2 id="define-evals-with-defineeval" tabindex="-1">Define evals with <code>defineEval</code> <a class="header-anchor" href="#define-evals-with-defineeval" aria-label="Permalink to &quot;Define evals with `defineEval`&quot;">​</a></h2><p>agentkit discovers evals under the project-root <code>evals/</code> directory, in <code>.eval.ts</code> or <code>.eval.js</code> files. That&#39;s a sibling of <code>agent/</code>, never inside it (<code>agent/evals/</code> is silently ignored). TypeScript is the normal authoring format.</p><p>The file path is the eval&#39;s identity, so you don&#39;t author an id. Directories group related evals: <code>evals/builds/api.eval.ts</code> becomes id <code>builds/api</code>. An <code>index</code> filename collapses to its directory, so <code>evals/builds/index.eval.ts</code> becomes <code>builds</code>.</p><p>An eval is a single <code>async test(t)</code>. You drive the agent with <code>t</code> and assert on the run with the same <code>t</code>:</p><div class="language-ts vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">ts</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span style="--shiki-light:#6A737D;--shiki-dark:#6A737D;">// evals/readiness.eval.ts</span></span>
22
- <span class="line"><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;">import</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;"> { defineEval, includes } </span><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;">from</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> &quot;@anysphere/agentkit/evals&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">;</span></span>
21
+ <div id="app"><div class="Layout" data-v-05085060><!--[--><!--]--><!--[--><span tabindex="-1" data-v-30dc752a></span><a href="#VPContent" class="VPSkipLink visually-hidden" data-v-30dc752a>Skip to content</a><!--]--><!----><header class="VPNav" data-v-05085060 data-v-cbe59881><div class="VPNavBar" data-v-cbe59881 data-v-dd8a1864><div class="wrapper" data-v-dd8a1864><div class="container" data-v-dd8a1864><div class="title" data-v-dd8a1864><div class="VPNavBarTitle has-sidebar" data-v-dd8a1864 data-v-f9346b2b><a class="title" href="/docs/" data-v-f9346b2b><!--[--><!--]--><!----><span data-v-f9346b2b>agentkit</span><!--[--><!--]--></a></div></div><div class="content" data-v-dd8a1864><div class="content-body" data-v-dd8a1864><!--[--><!--]--><div class="VPNavBarSearch search" data-v-dd8a1864><!--[--><!----><div id="local-search"><button type="button" class="DocSearch DocSearch-Button" aria-label="Search"><span class="DocSearch-Button-Container"><span class="vp-icon DocSearch-Search-Icon"></span><span class="DocSearch-Button-Placeholder">Search</span></span><span class="DocSearch-Button-Keys"><kbd class="DocSearch-Button-Key"></kbd><kbd class="DocSearch-Button-Key">K</kbd></span></button></div><!--]--></div><nav aria-labelledby="main-nav-aria-label" class="VPNavBarMenu menu" data-v-dd8a1864 data-v-7ed6e5ac><span id="main-nav-aria-label" class="visually-hidden" data-v-7ed6e5ac> Main Navigation </span><!--[--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/quickstart.html" tabindex="0" data-v-7ed6e5ac data-v-65720067><!--[--><span data-v-65720067>Quickstart</span><!--]--></a><!--]--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/guides/webhooks.html" tabindex="0" data-v-7ed6e5ac data-v-65720067><!--[--><span data-v-65720067>Guides</span><!--]--></a><!--]--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/reference/project-layout.html" tabindex="0" data-v-7ed6e5ac data-v-65720067><!--[--><span data-v-65720067>Reference</span><!--]--></a><!--]--><!--]--></nav><!----><div class="VPNavBarAppearance appearance" data-v-dd8a1864 data-v-2b2b395e><button class="VPSwitch VPSwitchAppearance" type="button" role="switch" title aria-checked="false" data-v-2b2b395e data-v-182deabf data-v-2562b538><span class="check" data-v-2562b538><span class="icon" data-v-2562b538><!--[--><span class="vpi-sun sun" data-v-182deabf></span><span class="vpi-moon moon" data-v-182deabf></span><!--]--></span></span></button></div><!----><div class="VPFlyout VPNavBarExtra extra" data-v-dd8a1864 data-v-f3b584b7 data-v-82133de6><button type="button" class="button" aria-haspopup="true" aria-expanded="false" aria-label="extra navigation" data-v-82133de6><span class="vpi-more-horizontal icon" data-v-82133de6></span></button><div class="menu" data-v-82133de6><div class="VPMenu" data-v-82133de6 data-v-759c9e19><!----><!--[--><!--[--><!----><div class="group" data-v-f3b584b7><div class="item appearance" data-v-f3b584b7><p class="label" data-v-f3b584b7>Appearance</p><div class="appearance-action" data-v-f3b584b7><button class="VPSwitch VPSwitchAppearance" type="button" role="switch" title aria-checked="false" data-v-f3b584b7 data-v-182deabf data-v-2562b538><span class="check" data-v-2562b538><span class="icon" data-v-2562b538><!--[--><span class="vpi-sun sun" data-v-182deabf></span><span class="vpi-moon moon" data-v-182deabf></span><!--]--></span></span></button></div></div></div><!----><!--]--><!--]--></div></div></div><!--[--><!--]--><button type="button" class="VPNavBarHamburger hamburger" aria-label="mobile navigation" aria-expanded="false" aria-controls="VPNavScreen" data-v-dd8a1864 data-v-ab07e844><span class="container" data-v-ab07e844><span class="top" data-v-ab07e844></span><span class="middle" data-v-ab07e844></span><span class="bottom" data-v-ab07e844></span></span></button></div></div></div></div><div class="divider" data-v-dd8a1864><div class="divider-line" data-v-dd8a1864></div></div></div><!----></header><div class="VPLocalNav has-sidebar empty" data-v-05085060 data-v-826b8761><div class="container" data-v-826b8761><button class="menu" aria-expanded="false" aria-controls="VPSidebarNav" data-v-826b8761><span class="vpi-align-left menu-icon" data-v-826b8761></span><span class="menu-text" data-v-826b8761>Menu</span></button><div class="VPLocalNavOutlineDropdown" style="--vp-vh:0px;" data-v-826b8761 data-v-01af9e73><button data-v-01af9e73>Return to top</button><!----></div></div></div><aside class="VPSidebar" data-v-05085060 data-v-5cd5b4d8><div class="curtain" data-v-5cd5b4d8></div><nav class="nav" id="VPSidebarNav" aria-labelledby="sidebar-aria-label" tabindex="-1" data-v-5cd5b4d8><span class="visually-hidden" id="sidebar-aria-label" data-v-5cd5b4d8> Sidebar Navigation </span><!--[--><!--]--><!--[--><div class="no-transition group" data-v-6b69f472><section class="VPSidebarItem level-0" data-v-6b69f472 data-v-b6893b1b><div class="item" role="button" tabindex="0" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><h2 class="text" data-v-b6893b1b>Getting started</h2><!----></div><div class="items" data-v-b6893b1b><!--[--><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Overview</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/quickstart.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Quickstart</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/scaffolding-agents.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Scaffold an agent with Cursor</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/concepts.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Concepts</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-6b69f472><section class="VPSidebarItem level-0 has-active" data-v-6b69f472 data-v-b6893b1b><div class="item" role="button" tabindex="0" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><h2 class="text" data-v-b6893b1b>Improving agents</h2><!----></div><div class="items" data-v-b6893b1b><!--[--><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/building-with-agents.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Building agents with agents</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/evals.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Evals</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/ab.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Live A/B metrics</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/storage.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Storage</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/hillclimbing.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Hillclimbing</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-6b69f472><section class="VPSidebarItem level-0" data-v-6b69f472 data-v-b6893b1b><div class="item" role="button" tabindex="0" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><h2 class="text" data-v-b6893b1b>Guides</h2><!----></div><div class="items" data-v-b6893b1b><!--[--><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/guides/webhooks.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Webhooks & custom channels</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/guides/github.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>GitHub</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/guides/slack.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Slack</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/guides/human-in-the-loop.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Human-in-the-loop approvals</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/guides/agent-to-agent.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Agent-to-agent</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/guides/cloud-runtime.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Cloud runtime</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-6b69f472><section class="VPSidebarItem level-0" data-v-6b69f472 data-v-b6893b1b><div class="item" role="button" tabindex="0" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><h2 class="text" data-v-b6893b1b>Operating</h2><!----></div><div class="items" data-v-b6893b1b><!--[--><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/deployment.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Deployment</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/troubleshooting.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Fix common problems</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-6b69f472><section class="VPSidebarItem level-0 collapsible" data-v-6b69f472 data-v-b6893b1b><div class="item" role="button" tabindex="0" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><h2 class="text" data-v-b6893b1b>Reference</h2><div class="caret" role="button" aria-label="toggle section" tabindex="0" data-v-b6893b1b><span class="vpi-chevron-right caret-icon" data-v-b6893b1b></span></div></div><div class="items" data-v-b6893b1b><!--[--><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/project-layout.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Project layout</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/agent-config.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Agent config</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/instructions.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Instructions</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/tools.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Tools</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/skills.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Skills</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/connections.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>MCP Connections</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/subagents.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Subagents</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/channels.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Channels</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/schedules.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Schedules & reminders</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/hooks.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Hooks</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/sessions.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Sessions & streaming</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/playground.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Playground</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/cli.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>CLI</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/http-api.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>HTTP API</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><!--]--><!--[--><!--]--></nav></aside><div class="VPContent has-sidebar" id="VPContent" data-v-05085060 data-v-8eedf2ac><div class="VPDoc has-sidebar has-aside" data-v-8eedf2ac data-v-53b9fda6><!--[--><!--]--><div class="container" data-v-53b9fda6><div class="aside" data-v-53b9fda6><div class="aside-curtain" data-v-53b9fda6></div><div class="aside-container" data-v-53b9fda6><div class="aside-content" data-v-53b9fda6><div class="VPDocAside" data-v-53b9fda6 data-v-e448ecab><!--[--><!--]--><!--[--><!--]--><nav aria-labelledby="doc-outline-aria-label" class="VPDocAsideOutline" data-v-e448ecab data-v-b0b42374><div class="content" data-v-b0b42374><div class="outline-marker" data-v-b0b42374></div><div aria-level="2" class="outline-title" id="doc-outline-aria-label" role="heading" data-v-b0b42374>On this page</div><ul class="VPDocOutlineItem root" data-v-b0b42374 data-v-ae03bd52><!--[--><!--]--></ul></div></nav><!--[--><!--]--><div class="spacer" data-v-e448ecab></div><!--[--><!--]--><!----><!--[--><!--]--><!--[--><!--]--></div></div></div></div><div class="content" data-v-53b9fda6><div class="content-container" data-v-53b9fda6><!--[--><!--]--><main class="main" data-v-53b9fda6><div style="position:relative;" class="vp-doc _docs_evals" data-v-53b9fda6><div><h1 id="evals" tabindex="-1">Evals <a class="header-anchor" href="#evals" aria-label="Permalink to &quot;Evals&quot;">​</a></h1><p>An eval is a repeatable check that runs your agent against a fixed input and gates the recorded trajectory: the run completed, the right tool ran, the reply has the right shape. Evals are how you know a prompt tweak helped, a refactor didn&#39;t regress the agent, and last month&#39;s fix is still holding.</p><p>Evals exercise the same surface your users hit. The runner starts (or targets) a real agent server, drives sessions over the public API, and grades what comes back. A passing eval means the agent started, accepted a message, and did what you asserted.</p><div class="note custom-block github-alert"><p class="custom-block-title">NOTE</p><p>Import paths here use <code>@cursor/july/evals</code>. On projects still using <code>@cursor/july</code>, swap the import and run <code>agent-serve eval</code>. See <a href="./README.html#run-the-cli">Run the CLI</a> for the full rename table.</p></div><h2 id="define-evals-with-defineeval" tabindex="-1">Define evals with <code>defineEval</code> <a class="header-anchor" href="#define-evals-with-defineeval" aria-label="Permalink to &quot;Define evals with `defineEval`&quot;">​</a></h2><p>agentkit discovers evals under the project-root <code>evals/</code> directory, in <code>.eval.ts</code> or <code>.eval.js</code> files. That&#39;s a sibling of <code>agent/</code>, never inside it (<code>agent/evals/</code> is silently ignored). TypeScript is the normal authoring format.</p><p>The file path is the eval&#39;s identity, so you don&#39;t author an id. Directories group related evals: <code>evals/builds/api.eval.ts</code> becomes id <code>builds/api</code>. An <code>index</code> filename collapses to its directory, so <code>evals/builds/index.eval.ts</code> becomes <code>builds</code>.</p><p>An eval is a single <code>async test(t)</code>. You drive the agent with <code>t</code> and assert on the run with the same <code>t</code>:</p><div class="language-ts vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">ts</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span style="--shiki-light:#6A737D;--shiki-dark:#6A737D;">// evals/readiness.eval.ts</span></span>
22
+ <span class="line"><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;">import</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;"> { defineEval, includes } </span><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;">from</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> &quot;@cursor/july/evals&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">;</span></span>
23
23
  <span class="line"></span>
24
24
  <span class="line"><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;">export</span><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;"> default</span><span style="--shiki-light:#6F42C1;--shiki-dark:#B392F0;"> defineEval</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">({</span></span>
25
25
  <span class="line"><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;"> description: </span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;">&quot;Inspects a PR without approving it.&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">,</span></span>
@@ -63,7 +63,7 @@
63
63
  <span class="line"><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">});</span></span></code></pre></div><p>Case ids must be single path segments, unique within the file. Each case can set its own <code>description</code>, <code>tags</code>, and <code>timeoutMs</code>. A case-level value replaces the file-level value for that datapoint.</p><h2 id="configure-eval-runs" tabindex="-1">Configure eval runs <a class="header-anchor" href="#configure-eval-runs" aria-label="Permalink to &quot;Configure eval runs&quot;">​</a></h2><p>Each project with evals needs <code>evals/evals.config.ts</code> or <code>evals/evals.config.js</code>, and it must set <code>maxConcurrency</code>. Each case issues real model-provider requests, so concurrency is capped hard at 200. Existing projects use 20. Discovery with <code>eval --list</code> works without this file, but running a case does not.</p><div class="language-ts vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">ts</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;">import</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;"> {</span></span>
64
64
  <span class="line"><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;"> defineEvalConfig,</span></span>
65
65
  <span class="line"><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;"> persistEvalRunsToDir,</span></span>
66
- <span class="line"><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">} </span><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;">from</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> &quot;@anysphere/agent-serve/evals&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">;</span></span>
66
+ <span class="line"><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">} </span><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;">from</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> &quot;@cursor/july/evals&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">;</span></span>
67
67
  <span class="line"></span>
68
68
  <span class="line"><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;">export</span><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;"> default</span><span style="--shiki-light:#6F42C1;--shiki-dark:#B392F0;"> defineEvalConfig</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">({</span></span>
69
69
  <span class="line"><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;"> maxConcurrency: </span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">20</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">, </span><span style="--shiki-light:#6A737D;--shiki-dark:#6A737D;">// required</span></span>
@@ -103,7 +103,7 @@
103
103
  <span class="line"><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;"> }</span></span>
104
104
  <span class="line"><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;"> ]</span></span>
105
105
  <span class="line"><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">}</span></span></code></pre></div><p>Each case result can also include <code>description</code>, <code>finalText</code>, <code>tools</code>, <code>error</code>, and tool arguments or output. This shape lets CI report the failed assertion without parsing terminal text.</p><h2 id="run-evals-in-the-playground" tabindex="-1">Run evals in the playground <a class="header-anchor" href="#run-evals-in-the-playground" aria-label="Permalink to &quot;Run evals in the playground&quot;">​</a></h2><p>Start the server with <code>--dev</code>, open the playground, and choose <strong>Evals</strong>. You can run every case or one case, watch progress, and open the resulting session trace.</p><div class="language-bash vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">bash</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span style="--shiki-light:#6F42C1;--shiki-dark:#B392F0;">agentkit</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> serve</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> --dir</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> .</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> --dev</span></span></code></pre></div><p>Playground runs target the live server instead of an ephemeral one. Their sessions appear in the session list. One eval batch can run at a time. By default those batches are <strong>in-memory only</strong> (capped by <code>maxPlaygroundRuns</code>); set <code>persistRuns</code> in <code>evals.config.ts</code> if you need them after a serve restart — see <a href="#configure-eval-runs">Configure eval runs</a>.</p><p>The UI uses the playground eval routes (available without <code>--dev</code>): <code>GET /v1/dev/evals</code> lists datapoints and config (includes <code>maxPlaygroundRuns</code> / <code>durableRuns</code>), <code>GET /v1/dev/evals/runs</code> rehydrates recent batches after navigation, <code>POST /v1/dev/evals/runs</code> starts a batch, and <code>GET /v1/dev/evals/runs/:runId</code> polls it. See <a href="./reference/http-api.html#playground-eval-routes">Playground eval routes</a>. The start request returns <code>202</code> while cases run in the background. Poll until the snapshot status becomes <code>completed</code> or <code>failed</code>. Configuration errors appear on a failed snapshot.</p><h2 id="what-good-cases-assert" tabindex="-1">What good cases assert <a class="header-anchor" href="#what-good-cases-assert" aria-label="Permalink to &quot;What good cases assert&quot;">​</a></h2><p>Gate decisions and shape, not prose. Model wording varies run to run. Tool choice, tool avoidance, and output structure are the stable contract.</p><ol><li><code>t.succeeded()</code>: always, first.</li><li>The tool decision: <code>calledTool</code> for the intended path, <code>notCalledTool</code> for the likely wrong alternative. The pair is stronger than either alone.</li><li>Output shape: a regex for the contract (<code>/ready|blocked/i</code>, a JSON marker, a findings-block fence), never exact sentences.</li><li>For structured output, parse <code>t.reply</code> and check fields with <code>satisfies</code> instead of substring-matching JSON.</li></ol><p>The common failure modes: asserting exact phrasing, packing more than about five gates into one case (split it), and cases that depend on live external state that drifts (pin the input; see fixtures).</p><h2 id="pick-fixtures-by-agent-type" tabindex="-1">Pick fixtures by agent type <a class="header-anchor" href="#pick-fixtures-by-agent-type" aria-label="Permalink to &quot;Pick fixtures by agent type&quot;">​</a></h2><p>The right fixture depends on the surface under test.</p><table tabindex="0"><thead><tr><th>Agent surface</th><th>Fixture</th></tr></thead><tbody><tr><td>Chat / domain assistant</td><td>A canonical prompt string, chosen once and frozen</td></tr><tr><td>Tool-heavy</td><td>Run <code>agentkit call &lt;tool&gt;</code> first to pin what the tool returns, then freeze the prompt that triggers it</td></tr><tr><td>GitHub webhook</td><td><code>agentkit github replay &lt;pr&gt; --events &#39;*&#39; --dry-run --out fixtures/github</code> snapshots real payloads for offline replay (<a href="./guides/github.html">GitHub guide</a>)</td></tr><tr><td>PR reviewer with host preparation</td><td>Diff, metadata, and gold labels pinned to commit SHAs; keep any live PR matrix small</td></tr><tr><td>Workspace-dependent</td><td><code>workspaceFiles</code> in <code>t.send</code> options, never developer-machine paths</td></tr></tbody></table><p>Tag the fast, reliably passing core <code>smoke</code> and run <code>--tag smoke</code> in the inner loop. Leave slow or flaky-prone cases untagged for explicit runs.</p><h3 id="materialize-api-backed-fixtures" tabindex="-1">Materialize API-backed fixtures <a class="header-anchor" href="#materialize-api-backed-fixtures" aria-label="Permalink to &quot;Materialize API-backed fixtures&quot;">​</a></h3><p>An input that only points at external data, such as a pull request URL, snapshot id, or pair of commit SHAs, is not self-contained. Fetch it once and commit the rendered fixture before you expand the suite.</p><ol><li>Save the diff, metadata, and labels under <code>fixtures/</code> at pinned revisions.</li><li>Seed those files with <code>workspaceFiles</code>, or read them from the fixture directory.</li><li>Assert decisions and output shape against the saved evidence.</li><li>Keep a small <code>smoke</code> subset for any remaining live pipeline checks.</li></ol><p><code>maxConcurrency</code> limits parallel datapoints. It does not limit model or API fan-out inside one datapoint. Materialized fixtures prevent a large suite from exhausting provider and GitHub rate limits. The <a href="./../skills/evals/SKILL.html">evals skill</a> has the full fixture workflow.</p><h2 id="keep-improvements-with-regression-evals" tabindex="-1">Keep improvements with regression evals <a class="header-anchor" href="#keep-improvements-with-regression-evals" aria-label="Permalink to &quot;Keep improvements with regression evals&quot;">​</a></h2><p>Every <a href="./hillclimbing.html">hillclimb</a> round that keeps a change must land an eval that would have failed before the change. If you can&#39;t express the improvement as a gate (a <code>calledTool</code> shift, a bounded <code>action.result</code> count, an output-shape regex), the improvement is unverified, and it&#39;ll regress silently.</p><p>The rule cuts the other way too: never weaken an existing gate to make a round pass. That&#39;s the freeze line moving, and it turns your regression suite into a list of checks that no longer protect anything.</p><h2 id="compare-variants-on-live-traffic" tabindex="-1">Compare variants on live traffic <a class="header-anchor" href="#compare-variants-on-live-traffic" aria-label="Permalink to &quot;Compare variants on live traffic&quot;">​</a></h2><p>Use <code>defineAB</code> to compare variant metrics on live sessions. It is not a test runner and has no <code>agentkit ab</code> command. Keep <code>defineEval</code> as the regression ratchet. Eval sessions do not enroll or change live metrics. See <a href="./ab.html">Live A/B metrics</a> for assignment, behavior, collection, and inspection.</p><h2 id="what-s-next" tabindex="-1">What&#39;s next <a class="header-anchor" href="#what-s-next" aria-label="Permalink to &quot;What&#39;s next&quot;">​</a></h2><p>Continue with these pages:</p><ul><li><a href="./ab.html">Live A/B metrics</a>: sticky variants and cumulative metrics on live sessions</li><li><a href="./hillclimbing.html">Hillclimbing</a>: the loop evals make trustworthy</li><li><a href="./building-with-agents.html">Building agents with agents</a>: have a coding agent write the first suite</li><li><a href="./guides/github.html">GitHub guide</a>: deterministic webhook fixtures with <code>github replay</code></li><li><a href="./reference/sessions.html">Sessions and streaming</a>: the events <code>t.events</code> contains</li></ul></div></div></main><footer class="VPDocFooter" data-v-53b9fda6 data-v-3958cc3f><!--[--><!--]--><!----><nav class="prev-next" aria-labelledby="doc-footer-aria-label" data-v-3958cc3f><span class="visually-hidden" id="doc-footer-aria-label" data-v-3958cc3f>Pager</span><div class="pager" data-v-3958cc3f><a class="VPLink link pager-link prev" href="/docs/building-with-agents.html" data-v-3958cc3f><!--[--><span class="desc" data-v-3958cc3f>Previous page</span><span class="title" data-v-3958cc3f>Building agents with agents</span><!--]--></a></div><div class="pager" data-v-3958cc3f><a class="VPLink link pager-link next" href="/docs/ab.html" data-v-3958cc3f><!--[--><span class="desc" data-v-3958cc3f>Next page</span><span class="title" data-v-3958cc3f>Live A/B metrics</span><!--]--></a></div></nav></footer><!--[--><!--]--></div></div></div><!--[--><!--]--></div></div><!----><!--[--><!--]--></div></div>
106
- <script>window.__VP_HASH_MAP__=JSON.parse("{\"ab.md\":\"COdXkces\",\"building-with-agents.md\":\"CJCtZCyi\",\"concepts.md\":\"Cfb9b-k1\",\"deployment.md\":\"Dx1TYNk5\",\"evals.md\":\"DPZ_MAnI\",\"guides_agent-to-agent.md\":\"CrtrsySy\",\"guides_cloud-runtime.md\":\"CYlNMTNp\",\"guides_github.md\":\"DwbKhCeS\",\"guides_human-in-the-loop.md\":\"Dvuctx7s\",\"guides_slack.md\":\"bv41fHfW\",\"guides_webhooks.md\":\"hFTik3lf\",\"hillclimbing.md\":\"DHNast08\",\"index.md\":\"BPKcj5AI\",\"quickstart.md\":\"tVPiGK_L\",\"reference_agent-config.md\":\"Bpd7HQwf\",\"reference_channels.md\":\"D7JTR03W\",\"reference_cli.md\":\"DA730zCu\",\"reference_connections.md\":\"C3vNH_DE\",\"reference_hooks.md\":\"BCEc3MyM\",\"reference_http-api.md\":\"DBAahtdz\",\"reference_instructions.md\":\"BC05LEQ8\",\"reference_playground.md\":\"CPZhfYaO\",\"reference_project-layout.md\":\"Bd_CKtNS\",\"reference_schedules.md\":\"D7qijxLk\",\"reference_sessions.md\":\"DdjWabnN\",\"reference_skills.md\":\"VQnlBT3Q\",\"reference_subagents.md\":\"CIRAVcPK\",\"reference_tools.md\":\"DF5kwlt0\",\"scaffolding-agents.md\":\"CyYfWGdc\",\"troubleshooting.md\":\"DkovjS7h\"}");window.__VP_SITE_DATA__=JSON.parse("{\"lang\":\"en-US\",\"dir\":\"ltr\",\"title\":\"agentkit\",\"description\":\"Filesystem-first framework for building and serving Cursor agents.\",\"base\":\"/docs/\",\"head\":[],\"router\":{\"prefetchLinks\":true},\"appearance\":true,\"themeConfig\":{\"nav\":[{\"text\":\"Quickstart\",\"link\":\"/quickstart\"},{\"text\":\"Guides\",\"link\":\"/guides/webhooks\"},{\"text\":\"Reference\",\"link\":\"/reference/project-layout\"}],\"search\":{\"provider\":\"local\"},\"outline\":{\"level\":[2,3]},\"sidebar\":[{\"text\":\"Getting started\",\"items\":[{\"text\":\"Overview\",\"link\":\"/\"},{\"text\":\"Quickstart\",\"link\":\"/quickstart\"},{\"text\":\"Scaffold an agent with Cursor\",\"link\":\"/scaffolding-agents\"},{\"text\":\"Concepts\",\"link\":\"/concepts\"}]},{\"text\":\"Improving agents\",\"items\":[{\"text\":\"Building agents with agents\",\"link\":\"/building-with-agents\"},{\"text\":\"Evals\",\"link\":\"/evals\"},{\"text\":\"Live A/B metrics\",\"link\":\"/ab\"},{\"text\":\"Hillclimbing\",\"link\":\"/hillclimbing\"}]},{\"text\":\"Guides\",\"items\":[{\"text\":\"Webhooks & custom channels\",\"link\":\"/guides/webhooks\"},{\"text\":\"GitHub\",\"link\":\"/guides/github\"},{\"text\":\"Slack\",\"link\":\"/guides/slack\"},{\"text\":\"Human-in-the-loop approvals\",\"link\":\"/guides/human-in-the-loop\"},{\"text\":\"Agent-to-agent\",\"link\":\"/guides/agent-to-agent\"},{\"text\":\"Cloud runtime\",\"link\":\"/guides/cloud-runtime\"}]},{\"text\":\"Operating\",\"items\":[{\"text\":\"Deployment\",\"link\":\"/deployment\"},{\"text\":\"Fix common problems\",\"link\":\"/troubleshooting\"}]},{\"text\":\"Reference\",\"collapsed\":false,\"items\":[{\"text\":\"Project layout\",\"link\":\"/reference/project-layout\"},{\"text\":\"Agent config\",\"link\":\"/reference/agent-config\"},{\"text\":\"Instructions\",\"link\":\"/reference/instructions\"},{\"text\":\"Tools\",\"link\":\"/reference/tools\"},{\"text\":\"Skills\",\"link\":\"/reference/skills\"},{\"text\":\"MCP Connections\",\"link\":\"/reference/connections\"},{\"text\":\"Subagents\",\"link\":\"/reference/subagents\"},{\"text\":\"Channels\",\"link\":\"/reference/channels\"},{\"text\":\"Schedules & reminders\",\"link\":\"/reference/schedules\"},{\"text\":\"Hooks\",\"link\":\"/reference/hooks\"},{\"text\":\"Sessions & streaming\",\"link\":\"/reference/sessions\"},{\"text\":\"Playground\",\"link\":\"/reference/playground\"},{\"text\":\"CLI\",\"link\":\"/reference/cli\"},{\"text\":\"HTTP API\",\"link\":\"/reference/http-api\"}]}]},\"locales\":{},\"scrollOffset\":134,\"cleanUrls\":false}");</script>
106
+ <script>window.__VP_HASH_MAP__=JSON.parse("{\"ab.md\":\"BMCZ6Hd7\",\"building-with-agents.md\":\"CJCtZCyi\",\"concepts.md\":\"Cfb9b-k1\",\"deployment.md\":\"DTKwE15Z\",\"evals.md\":\"DAgEc_hL\",\"guides_agent-to-agent.md\":\"Bpzgq2Pq\",\"guides_cloud-runtime.md\":\"gVzabdQL\",\"guides_github.md\":\"DOOCpqsW\",\"guides_human-in-the-loop.md\":\"DlUqsp1S\",\"guides_slack.md\":\"CCwqHvSV\",\"guides_webhooks.md\":\"B1EswtUu\",\"hillclimbing.md\":\"DHNast08\",\"index.md\":\"m81y7TY7\",\"quickstart.md\":\"BU6Iwi_9\",\"reference_agent-config.md\":\"DrW2JUM8\",\"reference_channels.md\":\"DdmiKgqf\",\"reference_cli.md\":\"Bv6pOxcF\",\"reference_connections.md\":\"zaEYCLHT\",\"reference_hooks.md\":\"DyLVfE1O\",\"reference_http-api.md\":\"Dx_nmDG6\",\"reference_instructions.md\":\"CgoV-YEb\",\"reference_playground.md\":\"CPZhfYaO\",\"reference_project-layout.md\":\"Bd_CKtNS\",\"reference_schedules.md\":\"w_F2mXB6\",\"reference_sessions.md\":\"DdjWabnN\",\"reference_skills.md\":\"B_jHN7JL\",\"reference_subagents.md\":\"zWAMNfi1\",\"reference_tools.md\":\"CqgJroI0\",\"scaffolding-agents.md\":\"C3pTrmoE\",\"storage.md\":\"CVnInNiN\",\"troubleshooting.md\":\"DkovjS7h\"}");window.__VP_SITE_DATA__=JSON.parse("{\"lang\":\"en-US\",\"dir\":\"ltr\",\"title\":\"agentkit\",\"description\":\"Filesystem-first framework for building and serving Cursor agents.\",\"base\":\"/docs/\",\"head\":[],\"router\":{\"prefetchLinks\":true},\"appearance\":true,\"themeConfig\":{\"nav\":[{\"text\":\"Quickstart\",\"link\":\"/quickstart\"},{\"text\":\"Guides\",\"link\":\"/guides/webhooks\"},{\"text\":\"Reference\",\"link\":\"/reference/project-layout\"}],\"search\":{\"provider\":\"local\"},\"outline\":{\"level\":[2,3]},\"sidebar\":[{\"text\":\"Getting started\",\"items\":[{\"text\":\"Overview\",\"link\":\"/\"},{\"text\":\"Quickstart\",\"link\":\"/quickstart\"},{\"text\":\"Scaffold an agent with Cursor\",\"link\":\"/scaffolding-agents\"},{\"text\":\"Concepts\",\"link\":\"/concepts\"}]},{\"text\":\"Improving agents\",\"items\":[{\"text\":\"Building agents with agents\",\"link\":\"/building-with-agents\"},{\"text\":\"Evals\",\"link\":\"/evals\"},{\"text\":\"Live A/B metrics\",\"link\":\"/ab\"},{\"text\":\"Storage\",\"link\":\"/storage\"},{\"text\":\"Hillclimbing\",\"link\":\"/hillclimbing\"}]},{\"text\":\"Guides\",\"items\":[{\"text\":\"Webhooks & custom channels\",\"link\":\"/guides/webhooks\"},{\"text\":\"GitHub\",\"link\":\"/guides/github\"},{\"text\":\"Slack\",\"link\":\"/guides/slack\"},{\"text\":\"Human-in-the-loop approvals\",\"link\":\"/guides/human-in-the-loop\"},{\"text\":\"Agent-to-agent\",\"link\":\"/guides/agent-to-agent\"},{\"text\":\"Cloud runtime\",\"link\":\"/guides/cloud-runtime\"}]},{\"text\":\"Operating\",\"items\":[{\"text\":\"Deployment\",\"link\":\"/deployment\"},{\"text\":\"Fix common problems\",\"link\":\"/troubleshooting\"}]},{\"text\":\"Reference\",\"collapsed\":false,\"items\":[{\"text\":\"Project layout\",\"link\":\"/reference/project-layout\"},{\"text\":\"Agent config\",\"link\":\"/reference/agent-config\"},{\"text\":\"Instructions\",\"link\":\"/reference/instructions\"},{\"text\":\"Tools\",\"link\":\"/reference/tools\"},{\"text\":\"Skills\",\"link\":\"/reference/skills\"},{\"text\":\"MCP Connections\",\"link\":\"/reference/connections\"},{\"text\":\"Subagents\",\"link\":\"/reference/subagents\"},{\"text\":\"Channels\",\"link\":\"/reference/channels\"},{\"text\":\"Schedules & reminders\",\"link\":\"/reference/schedules\"},{\"text\":\"Hooks\",\"link\":\"/reference/hooks\"},{\"text\":\"Sessions & streaming\",\"link\":\"/reference/sessions\"},{\"text\":\"Playground\",\"link\":\"/reference/playground\"},{\"text\":\"CLI\",\"link\":\"/reference/cli\"},{\"text\":\"HTTP API\",\"link\":\"/reference/http-api\"}]}]},\"locales\":{},\"scrollOffset\":134,\"cleanUrls\":false}");</script>
107
107
 
108
108
  </body>
109
109
  </html>
@@ -9,17 +9,17 @@
9
9
  <link rel="preload stylesheet" href="/docs/assets/style.Bj-IXa5W.css" as="style">
10
10
  <link rel="preload stylesheet" href="/docs/vp-icons.css" as="style">
11
11
 
12
- <script type="module" src="/docs/assets/app.DqfFEmJd.js"></script>
12
+ <script type="module" src="/docs/assets/app.BR0RbIdx.js"></script>
13
13
  <link rel="preload" href="/docs/assets/inter-roman-latin.Di8DUHzh.woff2" as="font" type="font/woff2" crossorigin="">
14
- <link rel="modulepreload" href="/docs/assets/chunks/theme.CZRvu_0q.js">
14
+ <link rel="modulepreload" href="/docs/assets/chunks/theme.2Kg8jIp_.js">
15
15
  <link rel="modulepreload" href="/docs/assets/chunks/framework.CAZyNGu9.js">
16
- <link rel="modulepreload" href="/docs/assets/guides_agent-to-agent.md.CrtrsySy.lean.js">
16
+ <link rel="modulepreload" href="/docs/assets/guides_agent-to-agent.md.Bpzgq2Pq.lean.js">
17
17
  <script id="check-dark-mode">(()=>{const e=localStorage.getItem("vitepress-theme-appearance")||"auto",a=window.matchMedia("(prefers-color-scheme: dark)").matches;(!e||e==="auto"?a:e==="dark")&&document.documentElement.classList.add("dark")})();</script>
18
18
  <script id="check-mac-os">document.documentElement.classList.toggle("mac",/Mac|iPhone|iPod|iPad/i.test(navigator.platform));</script>
19
19
  </head>
20
20
  <body>
21
- <div id="app"><div class="Layout" data-v-05085060><!--[--><!--]--><!--[--><span tabindex="-1" data-v-30dc752a></span><a href="#VPContent" class="VPSkipLink visually-hidden" data-v-30dc752a>Skip to content</a><!--]--><!----><header class="VPNav" data-v-05085060 data-v-cbe59881><div class="VPNavBar" data-v-cbe59881 data-v-dd8a1864><div class="wrapper" data-v-dd8a1864><div class="container" data-v-dd8a1864><div class="title" data-v-dd8a1864><div class="VPNavBarTitle has-sidebar" data-v-dd8a1864 data-v-f9346b2b><a class="title" href="/docs/" data-v-f9346b2b><!--[--><!--]--><!----><span data-v-f9346b2b>agentkit</span><!--[--><!--]--></a></div></div><div class="content" data-v-dd8a1864><div class="content-body" data-v-dd8a1864><!--[--><!--]--><div class="VPNavBarSearch search" data-v-dd8a1864><!--[--><!----><div id="local-search"><button type="button" class="DocSearch DocSearch-Button" aria-label="Search"><span class="DocSearch-Button-Container"><span class="vp-icon DocSearch-Search-Icon"></span><span class="DocSearch-Button-Placeholder">Search</span></span><span class="DocSearch-Button-Keys"><kbd class="DocSearch-Button-Key"></kbd><kbd class="DocSearch-Button-Key">K</kbd></span></button></div><!--]--></div><nav aria-labelledby="main-nav-aria-label" class="VPNavBarMenu menu" data-v-dd8a1864 data-v-7ed6e5ac><span id="main-nav-aria-label" class="visually-hidden" data-v-7ed6e5ac> Main Navigation </span><!--[--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/quickstart.html" tabindex="0" data-v-7ed6e5ac data-v-65720067><!--[--><span data-v-65720067>Quickstart</span><!--]--></a><!--]--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/guides/webhooks.html" tabindex="0" data-v-7ed6e5ac data-v-65720067><!--[--><span data-v-65720067>Guides</span><!--]--></a><!--]--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/reference/project-layout.html" tabindex="0" data-v-7ed6e5ac data-v-65720067><!--[--><span data-v-65720067>Reference</span><!--]--></a><!--]--><!--]--></nav><!----><div class="VPNavBarAppearance appearance" data-v-dd8a1864 data-v-2b2b395e><button class="VPSwitch VPSwitchAppearance" type="button" role="switch" title aria-checked="false" data-v-2b2b395e data-v-182deabf data-v-2562b538><span class="check" data-v-2562b538><span class="icon" data-v-2562b538><!--[--><span class="vpi-sun sun" data-v-182deabf></span><span class="vpi-moon moon" data-v-182deabf></span><!--]--></span></span></button></div><!----><div class="VPFlyout VPNavBarExtra extra" data-v-dd8a1864 data-v-f3b584b7 data-v-82133de6><button type="button" class="button" aria-haspopup="true" aria-expanded="false" aria-label="extra navigation" data-v-82133de6><span class="vpi-more-horizontal icon" data-v-82133de6></span></button><div class="menu" data-v-82133de6><div class="VPMenu" data-v-82133de6 data-v-759c9e19><!----><!--[--><!--[--><!----><div class="group" data-v-f3b584b7><div class="item appearance" data-v-f3b584b7><p class="label" data-v-f3b584b7>Appearance</p><div class="appearance-action" data-v-f3b584b7><button class="VPSwitch VPSwitchAppearance" type="button" role="switch" title aria-checked="false" data-v-f3b584b7 data-v-182deabf data-v-2562b538><span class="check" data-v-2562b538><span class="icon" data-v-2562b538><!--[--><span class="vpi-sun sun" data-v-182deabf></span><span class="vpi-moon moon" data-v-182deabf></span><!--]--></span></span></button></div></div></div><!----><!--]--><!--]--></div></div></div><!--[--><!--]--><button type="button" class="VPNavBarHamburger hamburger" aria-label="mobile navigation" aria-expanded="false" aria-controls="VPNavScreen" data-v-dd8a1864 data-v-ab07e844><span class="container" data-v-ab07e844><span class="top" data-v-ab07e844></span><span class="middle" data-v-ab07e844></span><span class="bottom" data-v-ab07e844></span></span></button></div></div></div></div><div class="divider" data-v-dd8a1864><div class="divider-line" data-v-dd8a1864></div></div></div><!----></header><div class="VPLocalNav has-sidebar empty" data-v-05085060 data-v-826b8761><div class="container" data-v-826b8761><button class="menu" aria-expanded="false" aria-controls="VPSidebarNav" data-v-826b8761><span class="vpi-align-left menu-icon" data-v-826b8761></span><span class="menu-text" data-v-826b8761>Menu</span></button><div class="VPLocalNavOutlineDropdown" style="--vp-vh:0px;" data-v-826b8761 data-v-01af9e73><button data-v-01af9e73>Return to top</button><!----></div></div></div><aside class="VPSidebar" data-v-05085060 data-v-5cd5b4d8><div class="curtain" data-v-5cd5b4d8></div><nav class="nav" id="VPSidebarNav" aria-labelledby="sidebar-aria-label" tabindex="-1" data-v-5cd5b4d8><span class="visually-hidden" id="sidebar-aria-label" data-v-5cd5b4d8> Sidebar Navigation </span><!--[--><!--]--><!--[--><div class="no-transition group" data-v-6b69f472><section class="VPSidebarItem level-0" data-v-6b69f472 data-v-b6893b1b><div class="item" role="button" tabindex="0" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><h2 class="text" data-v-b6893b1b>Getting started</h2><!----></div><div class="items" data-v-b6893b1b><!--[--><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Overview</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/quickstart.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Quickstart</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/scaffolding-agents.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Scaffold an agent with Cursor</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/concepts.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Concepts</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-6b69f472><section class="VPSidebarItem level-0" data-v-6b69f472 data-v-b6893b1b><div class="item" role="button" tabindex="0" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><h2 class="text" data-v-b6893b1b>Improving agents</h2><!----></div><div class="items" data-v-b6893b1b><!--[--><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/building-with-agents.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Building agents with agents</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/evals.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Evals</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/ab.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Live A/B metrics</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/hillclimbing.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Hillclimbing</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-6b69f472><section class="VPSidebarItem level-0 has-active" data-v-6b69f472 data-v-b6893b1b><div class="item" role="button" tabindex="0" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><h2 class="text" data-v-b6893b1b>Guides</h2><!----></div><div class="items" data-v-b6893b1b><!--[--><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/guides/webhooks.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Webhooks & custom channels</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/guides/github.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>GitHub</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/guides/slack.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Slack</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/guides/human-in-the-loop.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Human-in-the-loop approvals</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/guides/agent-to-agent.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Agent-to-agent</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/guides/cloud-runtime.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Cloud runtime</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-6b69f472><section class="VPSidebarItem level-0" data-v-6b69f472 data-v-b6893b1b><div class="item" role="button" tabindex="0" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><h2 class="text" data-v-b6893b1b>Operating</h2><!----></div><div class="items" data-v-b6893b1b><!--[--><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/deployment.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Deployment</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/troubleshooting.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Fix common problems</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-6b69f472><section class="VPSidebarItem level-0 collapsible" data-v-6b69f472 data-v-b6893b1b><div class="item" role="button" tabindex="0" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><h2 class="text" data-v-b6893b1b>Reference</h2><div class="caret" role="button" aria-label="toggle section" tabindex="0" data-v-b6893b1b><span class="vpi-chevron-right caret-icon" data-v-b6893b1b></span></div></div><div class="items" data-v-b6893b1b><!--[--><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/project-layout.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Project layout</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/agent-config.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Agent config</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/instructions.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Instructions</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/tools.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Tools</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/skills.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Skills</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/connections.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>MCP Connections</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/subagents.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Subagents</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/channels.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Channels</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/schedules.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Schedules & reminders</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/hooks.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Hooks</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/sessions.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Sessions & streaming</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/playground.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Playground</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/cli.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>CLI</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/http-api.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>HTTP API</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><!--]--><!--[--><!--]--></nav></aside><div class="VPContent has-sidebar" id="VPContent" data-v-05085060 data-v-8eedf2ac><div class="VPDoc has-sidebar has-aside" data-v-8eedf2ac data-v-53b9fda6><!--[--><!--]--><div class="container" data-v-53b9fda6><div class="aside" data-v-53b9fda6><div class="aside-curtain" data-v-53b9fda6></div><div class="aside-container" data-v-53b9fda6><div class="aside-content" data-v-53b9fda6><div class="VPDocAside" data-v-53b9fda6 data-v-e448ecab><!--[--><!--]--><!--[--><!--]--><nav aria-labelledby="doc-outline-aria-label" class="VPDocAsideOutline" data-v-e448ecab data-v-b0b42374><div class="content" data-v-b0b42374><div class="outline-marker" data-v-b0b42374></div><div aria-level="2" class="outline-title" id="doc-outline-aria-label" role="heading" data-v-b0b42374>On this page</div><ul class="VPDocOutlineItem root" data-v-b0b42374 data-v-ae03bd52><!--[--><!--]--></ul></div></nav><!--[--><!--]--><div class="spacer" data-v-e448ecab></div><!--[--><!--]--><!----><!--[--><!--]--><!--[--><!--]--></div></div></div></div><div class="content" data-v-53b9fda6><div class="content-container" data-v-53b9fda6><!--[--><!--]--><main class="main" data-v-53b9fda6><div style="position:relative;" class="vp-doc _docs_guides_agent-to-agent" data-v-53b9fda6><div><h1 id="agent-to-agent" tabindex="-1">Agent-to-agent <a class="header-anchor" href="#agent-to-agent" aria-label="Permalink to &quot;Agent-to-agent&quot;">​</a></h1><p>Every mounted agent is also an MCP server. So agents can delegate to each other without extra infrastructure. One agent hands a question to another, and the peer answers in its own session with its own instructions, tools, and context. Any external MCP client can do the same. A peer MCP connection makes the wiring one line.</p><p>This guide wires a <code>concierge</code> agent that delegates weather questions to a <code>weather-agent</code> peer mounted on the same host.</p><h2 id="the-mcp-endpoint" tabindex="-1">The MCP endpoint <a class="header-anchor" href="#the-mcp-endpoint" aria-label="Permalink to &quot;The MCP endpoint&quot;">​</a></h2><p>Each agent serves the Model Context Protocol over streamable HTTP at <code>/&lt;slug&gt;/v1/mcp</code> (or <code>/v1/mcp</code> in single mode). The surface is stateless (session identity travels in tool arguments) and runs the same route auth chain as the session API. It exposes three tools:</p><table tabindex="0"><thead><tr><th>Tool</th><th>Behavior</th></tr></thead><tbody><tr><td><code>ask</code></td><td>Send a message. Runs a model turn in this agent&#39;s own session and returns <code>{ status, sessionId, reply }</code>. Omit <code>sessionId</code> for a fresh session; pass it back to follow up.</td></tr><tr><td><code>check</code></td><td>Wait for or poll a running session (<code>waitSeconds: 0</code> polls without blocking).</td></tr><tr><td><code>call_tool</code></td><td>Call one of the agent&#39;s server tools directly, no model turn. Registered only when the agent has server tools.</td></tr></tbody></table><p>Waits are bounded at roughly 50 seconds, below typical MCP client request timeouts: a long turn returns <code>status: &quot;running&quot;</code> and the caller keeps waiting with <code>check</code>. Sessions created this way live on the <code>mcp</code> channel, bind to the calling principal, and show up in the playground and <code>GET /v1/sessions</code> like any other session.</p><p>Any MCP client can attach to this endpoint. It isn&#39;t only for other agentkit agents.</p><h2 id="wire-a-peer-mcp-connection" tabindex="-1">Wire a peer MCP connection <a class="header-anchor" href="#wire-a-peer-mcp-connection" aria-label="Permalink to &quot;Wire a peer MCP connection&quot;">​</a></h2><p>A peer MCP connection points one agent at another mounted on the same serve host. Author an MCP connection whose transport is the peer&#39;s slug:</p><div class="language-ts vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">ts</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span style="--shiki-light:#6A737D;--shiki-dark:#6A737D;">// agents/concierge/agent/mcp-connections/weather.ts</span></span>
22
- <span class="line"><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;">import</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;"> { defineConnection } </span><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;">from</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> &quot;@anysphere/agentkit/connections&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">;</span></span>
21
+ <div id="app"><div class="Layout" data-v-05085060><!--[--><!--]--><!--[--><span tabindex="-1" data-v-30dc752a></span><a href="#VPContent" class="VPSkipLink visually-hidden" data-v-30dc752a>Skip to content</a><!--]--><!----><header class="VPNav" data-v-05085060 data-v-cbe59881><div class="VPNavBar" data-v-cbe59881 data-v-dd8a1864><div class="wrapper" data-v-dd8a1864><div class="container" data-v-dd8a1864><div class="title" data-v-dd8a1864><div class="VPNavBarTitle has-sidebar" data-v-dd8a1864 data-v-f9346b2b><a class="title" href="/docs/" data-v-f9346b2b><!--[--><!--]--><!----><span data-v-f9346b2b>agentkit</span><!--[--><!--]--></a></div></div><div class="content" data-v-dd8a1864><div class="content-body" data-v-dd8a1864><!--[--><!--]--><div class="VPNavBarSearch search" data-v-dd8a1864><!--[--><!----><div id="local-search"><button type="button" class="DocSearch DocSearch-Button" aria-label="Search"><span class="DocSearch-Button-Container"><span class="vp-icon DocSearch-Search-Icon"></span><span class="DocSearch-Button-Placeholder">Search</span></span><span class="DocSearch-Button-Keys"><kbd class="DocSearch-Button-Key"></kbd><kbd class="DocSearch-Button-Key">K</kbd></span></button></div><!--]--></div><nav aria-labelledby="main-nav-aria-label" class="VPNavBarMenu menu" data-v-dd8a1864 data-v-7ed6e5ac><span id="main-nav-aria-label" class="visually-hidden" data-v-7ed6e5ac> Main Navigation </span><!--[--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/quickstart.html" tabindex="0" data-v-7ed6e5ac data-v-65720067><!--[--><span data-v-65720067>Quickstart</span><!--]--></a><!--]--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/guides/webhooks.html" tabindex="0" data-v-7ed6e5ac data-v-65720067><!--[--><span data-v-65720067>Guides</span><!--]--></a><!--]--><!--[--><a class="VPLink link VPNavBarMenuLink" href="/docs/reference/project-layout.html" tabindex="0" data-v-7ed6e5ac data-v-65720067><!--[--><span data-v-65720067>Reference</span><!--]--></a><!--]--><!--]--></nav><!----><div class="VPNavBarAppearance appearance" data-v-dd8a1864 data-v-2b2b395e><button class="VPSwitch VPSwitchAppearance" type="button" role="switch" title aria-checked="false" data-v-2b2b395e data-v-182deabf data-v-2562b538><span class="check" data-v-2562b538><span class="icon" data-v-2562b538><!--[--><span class="vpi-sun sun" data-v-182deabf></span><span class="vpi-moon moon" data-v-182deabf></span><!--]--></span></span></button></div><!----><div class="VPFlyout VPNavBarExtra extra" data-v-dd8a1864 data-v-f3b584b7 data-v-82133de6><button type="button" class="button" aria-haspopup="true" aria-expanded="false" aria-label="extra navigation" data-v-82133de6><span class="vpi-more-horizontal icon" data-v-82133de6></span></button><div class="menu" data-v-82133de6><div class="VPMenu" data-v-82133de6 data-v-759c9e19><!----><!--[--><!--[--><!----><div class="group" data-v-f3b584b7><div class="item appearance" data-v-f3b584b7><p class="label" data-v-f3b584b7>Appearance</p><div class="appearance-action" data-v-f3b584b7><button class="VPSwitch VPSwitchAppearance" type="button" role="switch" title aria-checked="false" data-v-f3b584b7 data-v-182deabf data-v-2562b538><span class="check" data-v-2562b538><span class="icon" data-v-2562b538><!--[--><span class="vpi-sun sun" data-v-182deabf></span><span class="vpi-moon moon" data-v-182deabf></span><!--]--></span></span></button></div></div></div><!----><!--]--><!--]--></div></div></div><!--[--><!--]--><button type="button" class="VPNavBarHamburger hamburger" aria-label="mobile navigation" aria-expanded="false" aria-controls="VPNavScreen" data-v-dd8a1864 data-v-ab07e844><span class="container" data-v-ab07e844><span class="top" data-v-ab07e844></span><span class="middle" data-v-ab07e844></span><span class="bottom" data-v-ab07e844></span></span></button></div></div></div></div><div class="divider" data-v-dd8a1864><div class="divider-line" data-v-dd8a1864></div></div></div><!----></header><div class="VPLocalNav has-sidebar empty" data-v-05085060 data-v-826b8761><div class="container" data-v-826b8761><button class="menu" aria-expanded="false" aria-controls="VPSidebarNav" data-v-826b8761><span class="vpi-align-left menu-icon" data-v-826b8761></span><span class="menu-text" data-v-826b8761>Menu</span></button><div class="VPLocalNavOutlineDropdown" style="--vp-vh:0px;" data-v-826b8761 data-v-01af9e73><button data-v-01af9e73>Return to top</button><!----></div></div></div><aside class="VPSidebar" data-v-05085060 data-v-5cd5b4d8><div class="curtain" data-v-5cd5b4d8></div><nav class="nav" id="VPSidebarNav" aria-labelledby="sidebar-aria-label" tabindex="-1" data-v-5cd5b4d8><span class="visually-hidden" id="sidebar-aria-label" data-v-5cd5b4d8> Sidebar Navigation </span><!--[--><!--]--><!--[--><div class="no-transition group" data-v-6b69f472><section class="VPSidebarItem level-0" data-v-6b69f472 data-v-b6893b1b><div class="item" role="button" tabindex="0" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><h2 class="text" data-v-b6893b1b>Getting started</h2><!----></div><div class="items" data-v-b6893b1b><!--[--><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Overview</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/quickstart.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Quickstart</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/scaffolding-agents.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Scaffold an agent with Cursor</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/concepts.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Concepts</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-6b69f472><section class="VPSidebarItem level-0" data-v-6b69f472 data-v-b6893b1b><div class="item" role="button" tabindex="0" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><h2 class="text" data-v-b6893b1b>Improving agents</h2><!----></div><div class="items" data-v-b6893b1b><!--[--><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/building-with-agents.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Building agents with agents</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/evals.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Evals</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/ab.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Live A/B metrics</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/storage.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Storage</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/hillclimbing.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Hillclimbing</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-6b69f472><section class="VPSidebarItem level-0 has-active" data-v-6b69f472 data-v-b6893b1b><div class="item" role="button" tabindex="0" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><h2 class="text" data-v-b6893b1b>Guides</h2><!----></div><div class="items" data-v-b6893b1b><!--[--><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/guides/webhooks.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Webhooks & custom channels</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/guides/github.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>GitHub</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/guides/slack.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Slack</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/guides/human-in-the-loop.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Human-in-the-loop approvals</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/guides/agent-to-agent.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Agent-to-agent</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/guides/cloud-runtime.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Cloud runtime</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-6b69f472><section class="VPSidebarItem level-0" data-v-6b69f472 data-v-b6893b1b><div class="item" role="button" tabindex="0" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><h2 class="text" data-v-b6893b1b>Operating</h2><!----></div><div class="items" data-v-b6893b1b><!--[--><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/deployment.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Deployment</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/troubleshooting.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Fix common problems</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><div class="no-transition group" data-v-6b69f472><section class="VPSidebarItem level-0 collapsible" data-v-6b69f472 data-v-b6893b1b><div class="item" role="button" tabindex="0" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><h2 class="text" data-v-b6893b1b>Reference</h2><div class="caret" role="button" aria-label="toggle section" tabindex="0" data-v-b6893b1b><span class="vpi-chevron-right caret-icon" data-v-b6893b1b></span></div></div><div class="items" data-v-b6893b1b><!--[--><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/project-layout.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Project layout</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/agent-config.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Agent config</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/instructions.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Instructions</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/tools.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Tools</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/skills.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Skills</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/connections.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>MCP Connections</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/subagents.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Subagents</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/channels.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Channels</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/schedules.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Schedules & reminders</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/hooks.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Hooks</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/sessions.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Sessions & streaming</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/playground.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>Playground</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/cli.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>CLI</p><!--]--></a><!----></div><!----></div><div class="VPSidebarItem level-1 is-link" data-v-b6893b1b><div class="item" data-v-b6893b1b><div class="indicator" data-v-b6893b1b></div><a class="VPLink link link" href="/docs/reference/http-api.html" data-v-b6893b1b><!--[--><p class="text" data-v-b6893b1b>HTTP API</p><!--]--></a><!----></div><!----></div><!--]--></div></section></div><!--]--><!--[--><!--]--></nav></aside><div class="VPContent has-sidebar" id="VPContent" data-v-05085060 data-v-8eedf2ac><div class="VPDoc has-sidebar has-aside" data-v-8eedf2ac data-v-53b9fda6><!--[--><!--]--><div class="container" data-v-53b9fda6><div class="aside" data-v-53b9fda6><div class="aside-curtain" data-v-53b9fda6></div><div class="aside-container" data-v-53b9fda6><div class="aside-content" data-v-53b9fda6><div class="VPDocAside" data-v-53b9fda6 data-v-e448ecab><!--[--><!--]--><!--[--><!--]--><nav aria-labelledby="doc-outline-aria-label" class="VPDocAsideOutline" data-v-e448ecab data-v-b0b42374><div class="content" data-v-b0b42374><div class="outline-marker" data-v-b0b42374></div><div aria-level="2" class="outline-title" id="doc-outline-aria-label" role="heading" data-v-b0b42374>On this page</div><ul class="VPDocOutlineItem root" data-v-b0b42374 data-v-ae03bd52><!--[--><!--]--></ul></div></nav><!--[--><!--]--><div class="spacer" data-v-e448ecab></div><!--[--><!--]--><!----><!--[--><!--]--><!--[--><!--]--></div></div></div></div><div class="content" data-v-53b9fda6><div class="content-container" data-v-53b9fda6><!--[--><!--]--><main class="main" data-v-53b9fda6><div style="position:relative;" class="vp-doc _docs_guides_agent-to-agent" data-v-53b9fda6><div><h1 id="agent-to-agent" tabindex="-1">Agent-to-agent <a class="header-anchor" href="#agent-to-agent" aria-label="Permalink to &quot;Agent-to-agent&quot;">​</a></h1><p>Every mounted agent is also an MCP server. So agents can delegate to each other without extra infrastructure. One agent hands a question to another, and the peer answers in its own session with its own instructions, tools, and context. Any external MCP client can do the same. A peer MCP connection makes the wiring one line.</p><p>This guide wires a <code>concierge</code> agent that delegates weather questions to a <code>weather-agent</code> peer mounted on the same host.</p><h2 id="the-mcp-endpoint" tabindex="-1">The MCP endpoint <a class="header-anchor" href="#the-mcp-endpoint" aria-label="Permalink to &quot;The MCP endpoint&quot;">​</a></h2><p>Each agent serves the Model Context Protocol over streamable HTTP at <code>/&lt;slug&gt;/v1/mcp</code> (or <code>/v1/mcp</code> in single mode). The surface is stateless (session identity travels in tool arguments) and runs the same route auth chain as the session API. It exposes three tools:</p><table tabindex="0"><thead><tr><th>Tool</th><th>Behavior</th></tr></thead><tbody><tr><td><code>ask</code></td><td>Send a message. Runs a model turn in this agent&#39;s own session and returns <code>{ status, sessionId, reply }</code>. Omit <code>sessionId</code> for a fresh session; pass it back to follow up.</td></tr><tr><td><code>check</code></td><td>Wait for or poll a running session (<code>waitSeconds: 0</code> polls without blocking).</td></tr><tr><td><code>call_tool</code></td><td>Call one of the agent&#39;s server tools directly, no model turn. Registered only when the agent has server tools.</td></tr></tbody></table><p>Waits are bounded at roughly 50 seconds, below typical MCP client request timeouts: a long turn returns <code>status: &quot;running&quot;</code> and the caller keeps waiting with <code>check</code>. Sessions created this way live on the <code>mcp</code> channel, bind to the calling principal, and show up in the playground and <code>GET /v1/sessions</code> like any other session.</p><p>Any MCP client can attach to this endpoint. It isn&#39;t only for other agentkit agents.</p><h2 id="wire-a-peer-mcp-connection" tabindex="-1">Wire a peer MCP connection <a class="header-anchor" href="#wire-a-peer-mcp-connection" aria-label="Permalink to &quot;Wire a peer MCP connection&quot;">​</a></h2><p>A peer MCP connection points one agent at another mounted on the same serve host. Author an MCP connection whose transport is the peer&#39;s slug:</p><div class="language-ts vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">ts</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span style="--shiki-light:#6A737D;--shiki-dark:#6A737D;">// agents/concierge/agent/mcp-connections/weather.ts</span></span>
22
+ <span class="line"><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;">import</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;"> { defineConnection } </span><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;">from</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> &quot;@cursor/july/connections&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">;</span></span>
23
23
  <span class="line"></span>
24
24
  <span class="line"><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;">export</span><span style="--shiki-light:#D73A49;--shiki-dark:#F97583;"> default</span><span style="--shiki-light:#6F42C1;--shiki-dark:#B392F0;"> defineConnection</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">({</span></span>
25
25
  <span class="line"><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;"> agent: </span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;">&quot;weather-agent&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">,</span></span>
@@ -28,7 +28,7 @@
28
28
  <span class="line"><span style="--shiki-light:#6F42C1;--shiki-dark:#B392F0;">agentkit</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> chat</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> --url</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> http://127.0.0.1:3000/concierge</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> \</span></span>
29
29
  <span class="line"><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;"> --message</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> &quot;What&#39;s the weather in Paris right now?&quot;</span></span>
30
30
  <span class="line"><span style="--shiki-light:#6A737D;--shiki-dark:#6A737D;"># concierge → weather.ask → weather-agent&#39;s own session/tools → reply</span></span></code></pre></div><h2 id="peer-or-subagent" tabindex="-1">Peer or subagent? <a class="header-anchor" href="#peer-or-subagent" aria-label="Permalink to &quot;Peer or subagent?&quot;">​</a></h2><p>The two delegation mechanisms solve different problems.</p><table tabindex="0"><thead><tr><th></th><th>Subagent (<code>agent/subagents/&lt;id&gt;/</code>)</th><th>Peer (<code>defineConnection({ agent })</code>)</th></tr></thead><tbody><tr><td>Runs as</td><td>an SDK custom subagent inside the parent&#39;s harness</td><td>an independent agent on the same host</td></tr><tr><td>Own tools, MCP connections, sessions</td><td>no, inherits the parent&#39;s surface</td><td>yes, everything is its own</td></tr><tr><td>Visible to others</td><td>only its parent</td><td>any MCP client, other agents, its own playground</td></tr><tr><td>Reach for it when</td><td>splitting one job into specialist roles</td><td>composing independently useful agents</td></tr></tbody></table><h2 id="how-peer-urls-resolve" tabindex="-1">How peer URLs resolve <a class="header-anchor" href="#how-peer-urls-resolve" aria-label="Permalink to &quot;How peer URLs resolve&quot;">​</a></h2><p>Peer URLs resolve when the server starts, so the ephemeral ports <code>run</code> and <code>eval</code> use work too. Local-runtime turns (and host-side <code>ctx.host.mcp</code> or channel-handler calls) reach the peer over loopback, which works under the default auth. Cloud-runtime turns execute on a VM that cannot reach this host&#39;s loopback address: pass <code>--public-url https://agentkit.example.com</code> (or <code>serve(dir, { publicUrl })</code>) so peers resolve to a reachable URL. Without one, peers are omitted from cloud turns and the server warns at startup. With <code>--bearer-token</code>, the token is attached to peer calls automatically so they pass the target agent&#39;s auth chain.</p><h2 id="guardrails" tabindex="-1">Guardrails <a class="header-anchor" href="#guardrails" aria-label="Permalink to &quot;Guardrails&quot;">​</a></h2><p>Unknown peer slugs and self-references fail <code>serve</code> at startup, so you find out immediately rather than at delegation time. Peers require the multi-agent layout (each agent mounted under its slug, the default).</p><p>The framework does not provide cross-host loop protection. If agent A&#39;s instructions delegate to B and B&#39;s delegate back to A, they can recurse. Scope each agent&#39;s delegation instructions narrowly. The concierge delegates <em>weather questions</em> to <code>weather-agent</code>, not everything.</p><h2 id="call-a-peer-from-host-code" tabindex="-1">Call a peer from host code <a class="header-anchor" href="#call-a-peer-from-host-code" aria-label="Permalink to &quot;Call a peer from host code&quot;">​</a></h2><p>Peer MCP connections are ordinary MCP connections, so deterministic host code can use them too. A channel handler or server tool can call <code>host.mcp.callTool(&quot;weather&quot;, &quot;ask&quot;, { message: &quot;…&quot; })</code> without any model turn deciding to. See <a href="./../reference/connections.html#every-mcp-connection-is-available-in-three-places">MCP connections</a> for the three places every MCP connection is available.</p><h2 id="what-s-next" tabindex="-1">What&#39;s next <a class="header-anchor" href="#what-s-next" aria-label="Permalink to &quot;What&#39;s next&quot;">​</a></h2><p>Continue with these pages:</p><ul><li><a href="./../reference/connections.html">MCP connections</a>: all four MCP connection transports</li><li><a href="./../reference/subagents.html">Subagents</a>: the in-harness alternative</li><li><a href="./../reference/http-api.html#mcp-endpoint">HTTP API</a>: the endpoint contract</li></ul></div></div></main><footer class="VPDocFooter" data-v-53b9fda6 data-v-3958cc3f><!--[--><!--]--><!----><nav class="prev-next" aria-labelledby="doc-footer-aria-label" data-v-3958cc3f><span class="visually-hidden" id="doc-footer-aria-label" data-v-3958cc3f>Pager</span><div class="pager" data-v-3958cc3f><a class="VPLink link pager-link prev" href="/docs/guides/human-in-the-loop.html" data-v-3958cc3f><!--[--><span class="desc" data-v-3958cc3f>Previous page</span><span class="title" data-v-3958cc3f>Human-in-the-loop approvals</span><!--]--></a></div><div class="pager" data-v-3958cc3f><a class="VPLink link pager-link next" href="/docs/guides/cloud-runtime.html" data-v-3958cc3f><!--[--><span class="desc" data-v-3958cc3f>Next page</span><span class="title" data-v-3958cc3f>Cloud runtime</span><!--]--></a></div></nav></footer><!--[--><!--]--></div></div></div><!--[--><!--]--></div></div><!----><!--[--><!--]--></div></div>
31
- <script>window.__VP_HASH_MAP__=JSON.parse("{\"ab.md\":\"COdXkces\",\"building-with-agents.md\":\"CJCtZCyi\",\"concepts.md\":\"Cfb9b-k1\",\"deployment.md\":\"Dx1TYNk5\",\"evals.md\":\"DPZ_MAnI\",\"guides_agent-to-agent.md\":\"CrtrsySy\",\"guides_cloud-runtime.md\":\"CYlNMTNp\",\"guides_github.md\":\"DwbKhCeS\",\"guides_human-in-the-loop.md\":\"Dvuctx7s\",\"guides_slack.md\":\"bv41fHfW\",\"guides_webhooks.md\":\"hFTik3lf\",\"hillclimbing.md\":\"DHNast08\",\"index.md\":\"BPKcj5AI\",\"quickstart.md\":\"tVPiGK_L\",\"reference_agent-config.md\":\"Bpd7HQwf\",\"reference_channels.md\":\"D7JTR03W\",\"reference_cli.md\":\"DA730zCu\",\"reference_connections.md\":\"C3vNH_DE\",\"reference_hooks.md\":\"BCEc3MyM\",\"reference_http-api.md\":\"DBAahtdz\",\"reference_instructions.md\":\"BC05LEQ8\",\"reference_playground.md\":\"CPZhfYaO\",\"reference_project-layout.md\":\"Bd_CKtNS\",\"reference_schedules.md\":\"D7qijxLk\",\"reference_sessions.md\":\"DdjWabnN\",\"reference_skills.md\":\"VQnlBT3Q\",\"reference_subagents.md\":\"CIRAVcPK\",\"reference_tools.md\":\"DF5kwlt0\",\"scaffolding-agents.md\":\"CyYfWGdc\",\"troubleshooting.md\":\"DkovjS7h\"}");window.__VP_SITE_DATA__=JSON.parse("{\"lang\":\"en-US\",\"dir\":\"ltr\",\"title\":\"agentkit\",\"description\":\"Filesystem-first framework for building and serving Cursor agents.\",\"base\":\"/docs/\",\"head\":[],\"router\":{\"prefetchLinks\":true},\"appearance\":true,\"themeConfig\":{\"nav\":[{\"text\":\"Quickstart\",\"link\":\"/quickstart\"},{\"text\":\"Guides\",\"link\":\"/guides/webhooks\"},{\"text\":\"Reference\",\"link\":\"/reference/project-layout\"}],\"search\":{\"provider\":\"local\"},\"outline\":{\"level\":[2,3]},\"sidebar\":[{\"text\":\"Getting started\",\"items\":[{\"text\":\"Overview\",\"link\":\"/\"},{\"text\":\"Quickstart\",\"link\":\"/quickstart\"},{\"text\":\"Scaffold an agent with Cursor\",\"link\":\"/scaffolding-agents\"},{\"text\":\"Concepts\",\"link\":\"/concepts\"}]},{\"text\":\"Improving agents\",\"items\":[{\"text\":\"Building agents with agents\",\"link\":\"/building-with-agents\"},{\"text\":\"Evals\",\"link\":\"/evals\"},{\"text\":\"Live A/B metrics\",\"link\":\"/ab\"},{\"text\":\"Hillclimbing\",\"link\":\"/hillclimbing\"}]},{\"text\":\"Guides\",\"items\":[{\"text\":\"Webhooks & custom channels\",\"link\":\"/guides/webhooks\"},{\"text\":\"GitHub\",\"link\":\"/guides/github\"},{\"text\":\"Slack\",\"link\":\"/guides/slack\"},{\"text\":\"Human-in-the-loop approvals\",\"link\":\"/guides/human-in-the-loop\"},{\"text\":\"Agent-to-agent\",\"link\":\"/guides/agent-to-agent\"},{\"text\":\"Cloud runtime\",\"link\":\"/guides/cloud-runtime\"}]},{\"text\":\"Operating\",\"items\":[{\"text\":\"Deployment\",\"link\":\"/deployment\"},{\"text\":\"Fix common problems\",\"link\":\"/troubleshooting\"}]},{\"text\":\"Reference\",\"collapsed\":false,\"items\":[{\"text\":\"Project layout\",\"link\":\"/reference/project-layout\"},{\"text\":\"Agent config\",\"link\":\"/reference/agent-config\"},{\"text\":\"Instructions\",\"link\":\"/reference/instructions\"},{\"text\":\"Tools\",\"link\":\"/reference/tools\"},{\"text\":\"Skills\",\"link\":\"/reference/skills\"},{\"text\":\"MCP Connections\",\"link\":\"/reference/connections\"},{\"text\":\"Subagents\",\"link\":\"/reference/subagents\"},{\"text\":\"Channels\",\"link\":\"/reference/channels\"},{\"text\":\"Schedules & reminders\",\"link\":\"/reference/schedules\"},{\"text\":\"Hooks\",\"link\":\"/reference/hooks\"},{\"text\":\"Sessions & streaming\",\"link\":\"/reference/sessions\"},{\"text\":\"Playground\",\"link\":\"/reference/playground\"},{\"text\":\"CLI\",\"link\":\"/reference/cli\"},{\"text\":\"HTTP API\",\"link\":\"/reference/http-api\"}]}]},\"locales\":{},\"scrollOffset\":134,\"cleanUrls\":false}");</script>
31
+ <script>window.__VP_HASH_MAP__=JSON.parse("{\"ab.md\":\"BMCZ6Hd7\",\"building-with-agents.md\":\"CJCtZCyi\",\"concepts.md\":\"Cfb9b-k1\",\"deployment.md\":\"DTKwE15Z\",\"evals.md\":\"DAgEc_hL\",\"guides_agent-to-agent.md\":\"Bpzgq2Pq\",\"guides_cloud-runtime.md\":\"gVzabdQL\",\"guides_github.md\":\"DOOCpqsW\",\"guides_human-in-the-loop.md\":\"DlUqsp1S\",\"guides_slack.md\":\"CCwqHvSV\",\"guides_webhooks.md\":\"B1EswtUu\",\"hillclimbing.md\":\"DHNast08\",\"index.md\":\"m81y7TY7\",\"quickstart.md\":\"BU6Iwi_9\",\"reference_agent-config.md\":\"DrW2JUM8\",\"reference_channels.md\":\"DdmiKgqf\",\"reference_cli.md\":\"Bv6pOxcF\",\"reference_connections.md\":\"zaEYCLHT\",\"reference_hooks.md\":\"DyLVfE1O\",\"reference_http-api.md\":\"Dx_nmDG6\",\"reference_instructions.md\":\"CgoV-YEb\",\"reference_playground.md\":\"CPZhfYaO\",\"reference_project-layout.md\":\"Bd_CKtNS\",\"reference_schedules.md\":\"w_F2mXB6\",\"reference_sessions.md\":\"DdjWabnN\",\"reference_skills.md\":\"B_jHN7JL\",\"reference_subagents.md\":\"zWAMNfi1\",\"reference_tools.md\":\"CqgJroI0\",\"scaffolding-agents.md\":\"C3pTrmoE\",\"storage.md\":\"CVnInNiN\",\"troubleshooting.md\":\"DkovjS7h\"}");window.__VP_SITE_DATA__=JSON.parse("{\"lang\":\"en-US\",\"dir\":\"ltr\",\"title\":\"agentkit\",\"description\":\"Filesystem-first framework for building and serving Cursor agents.\",\"base\":\"/docs/\",\"head\":[],\"router\":{\"prefetchLinks\":true},\"appearance\":true,\"themeConfig\":{\"nav\":[{\"text\":\"Quickstart\",\"link\":\"/quickstart\"},{\"text\":\"Guides\",\"link\":\"/guides/webhooks\"},{\"text\":\"Reference\",\"link\":\"/reference/project-layout\"}],\"search\":{\"provider\":\"local\"},\"outline\":{\"level\":[2,3]},\"sidebar\":[{\"text\":\"Getting started\",\"items\":[{\"text\":\"Overview\",\"link\":\"/\"},{\"text\":\"Quickstart\",\"link\":\"/quickstart\"},{\"text\":\"Scaffold an agent with Cursor\",\"link\":\"/scaffolding-agents\"},{\"text\":\"Concepts\",\"link\":\"/concepts\"}]},{\"text\":\"Improving agents\",\"items\":[{\"text\":\"Building agents with agents\",\"link\":\"/building-with-agents\"},{\"text\":\"Evals\",\"link\":\"/evals\"},{\"text\":\"Live A/B metrics\",\"link\":\"/ab\"},{\"text\":\"Storage\",\"link\":\"/storage\"},{\"text\":\"Hillclimbing\",\"link\":\"/hillclimbing\"}]},{\"text\":\"Guides\",\"items\":[{\"text\":\"Webhooks & custom channels\",\"link\":\"/guides/webhooks\"},{\"text\":\"GitHub\",\"link\":\"/guides/github\"},{\"text\":\"Slack\",\"link\":\"/guides/slack\"},{\"text\":\"Human-in-the-loop approvals\",\"link\":\"/guides/human-in-the-loop\"},{\"text\":\"Agent-to-agent\",\"link\":\"/guides/agent-to-agent\"},{\"text\":\"Cloud runtime\",\"link\":\"/guides/cloud-runtime\"}]},{\"text\":\"Operating\",\"items\":[{\"text\":\"Deployment\",\"link\":\"/deployment\"},{\"text\":\"Fix common problems\",\"link\":\"/troubleshooting\"}]},{\"text\":\"Reference\",\"collapsed\":false,\"items\":[{\"text\":\"Project layout\",\"link\":\"/reference/project-layout\"},{\"text\":\"Agent config\",\"link\":\"/reference/agent-config\"},{\"text\":\"Instructions\",\"link\":\"/reference/instructions\"},{\"text\":\"Tools\",\"link\":\"/reference/tools\"},{\"text\":\"Skills\",\"link\":\"/reference/skills\"},{\"text\":\"MCP Connections\",\"link\":\"/reference/connections\"},{\"text\":\"Subagents\",\"link\":\"/reference/subagents\"},{\"text\":\"Channels\",\"link\":\"/reference/channels\"},{\"text\":\"Schedules & reminders\",\"link\":\"/reference/schedules\"},{\"text\":\"Hooks\",\"link\":\"/reference/hooks\"},{\"text\":\"Sessions & streaming\",\"link\":\"/reference/sessions\"},{\"text\":\"Playground\",\"link\":\"/reference/playground\"},{\"text\":\"CLI\",\"link\":\"/reference/cli\"},{\"text\":\"HTTP API\",\"link\":\"/reference/http-api\"}]}]},\"locales\":{},\"scrollOffset\":134,\"cleanUrls\":false}");</script>
32
32
 
33
33
  </body>
34
34
  </html>