agents 0.29.0 → 0.30.1

package/README.md

@@ -509,8 +509,9 @@ export default {
 } satisfies ExportedHandler;
 ```
 
-Use `McpAgent`, `createLegacyMcpHandler`, and `WorkerTransport` from
-`agents/mcp` only when retaining Legacy session behavior.
+Install `@modelcontextprotocol/server` alongside `agents`. See
+[Creating MCP Servers](../../docs/agents/mcp-servers.md) for authentication,
+elicitation, and keeping state in bindings.
 
 ### Using MCP Tools
 

package/dist/browser/tools/pi.d.ts

@@ -73,7 +73,7 @@ type PiBrowserTool<TName extends string = "browser"> = ToolRegistration<
  *   readonly browser = new Browser({ provider: browserRun(this.env.BROWSER) });
  *   readonly registry = createRegistry();
  *   readonly harness = new PiHarness({
- *     harness: ({ storage, context }) => {
+ *     harness: ({ storage, context, onWake }) => {
  *       this.registry.install({
  *         name: "browser",
  *         tools: [
@@ -84,7 +84,11 @@ type PiBrowserTool<TName extends string = "browser"> = ToolRegistration<
  *           })
  *         ]
  *       });
- *       return Harness.open(storage, { models, registry: this.registry }, context);
+ *       return Harness.open(
+ *         storage,
+ *         { models, registry: this.registry, onWake },
+ *         context
+ *       );
  *     }
  *   });
  *   readonly lifecycle = Lifecycle.install(this)

package/dist/browser/tools/pi.js

@@ -78,7 +78,7 @@ function conversationScope(api) {
 *   readonly browser = new Browser({ provider: browserRun(this.env.BROWSER) });
 *   readonly registry = createRegistry();
 *   readonly harness = new PiHarness({
-*     harness: ({ storage, context }) => {
+*     harness: ({ storage, context, onWake }) => {
 *       this.registry.install({
 *         name: "browser",
 *         tools: [
@@ -89,7 +89,11 @@ function conversationScope(api) {
 *           })
 *         ]
 *       });
-*       return Harness.open(storage, { models, registry: this.registry }, context);
+*       return Harness.open(
+*         storage,
+*         { models, registry: this.registry, onWake },
+*         context
+*       );
 *     }
 *   });
 *   readonly lifecycle = Lifecycle.install(this)

package/dist/browser/tools/pi.js.map

@@ -1 +1 @@
-{"version":3,"file":"pi.js","names":[],"sources":["../../../src/browser/tools/pi.ts"],"sourcesContent":["import {\n  Type,\n  type ImageContent,\n  type TextContent\n} from \"@earendil-works/pi-ai\";\nimport type {\n  ToolExecutionApi,\n  ToolExecutionResult,\n  ToolRegistration\n} from \"@earendil-works/pi-durable\";\nimport {\n  createBrowserToolCore,\n  type BrowserToolOptions,\n  type BrowserToolOutput\n} from \"../browser-tool\";\nimport type { BrowserNewTab } from \"../session-connector\";\nimport {\n  browserExecuteModelOutput,\n  browserScreenshotOutput\n} from \"../tool-helpers\";\n\nexport type {\n  BrowserToolInput,\n  BrowserToolOptions,\n  BrowserToolOutput\n} from \"../browser-tool\";\nexport type { BrowserNewTab, BrowserSource } from \"../session-connector\";\n\nexport interface PiBrowserToolOptions<\n  TName extends string = \"browser\"\n> extends Omit<BrowserToolOptions, \"scope\"> {\n  /** The tool's name. Default `\"browser\"`. */\n  name?: TName;\n  /**\n   * Which calls share an active tab: a string for every call, or a function\n   * of the call. Defaults to `conversation:<id>` for the calling\n   * conversation, so each\n   * conversation (and each fork or subagent) on the Durable Object keeps its\n   * own tab, while cookies and logins stay shared.\n   */\n  scope?: string | ((api: ToolExecutionApi) => string);\n}\n\nconst browserToolParameters = Type.Object({\n  code: Type.String({\n    description: \"An async arrow function that drives the browser with `cdp`\"\n  })\n});\n\n/**\n * What a UI can show about a run without reading the content: the run's\n * outcome and what happened to the browser.\n */\nexport type PiBrowserToolDetails = {\n  executionId: string;\n  status: BrowserToolOutput[\"status\"];\n  restarted?: true;\n  newTabs?: { targetId: string; url?: string; title?: string }[];\n};\n\nexport type PiBrowserTool<TName extends string = \"browser\"> = ToolRegistration<\n  typeof browserToolParameters,\n  PiBrowserToolDetails\n> & {\n  readonly name: TName;\n};\n\nfunction details(output: BrowserToolOutput): PiBrowserToolDetails {\n  return {\n    executionId: output.executionId,\n    status: output.status,\n    ...(output.restarted ? { restarted: true as const } : {}),\n    ...(output.newTabs\n      ? {\n          newTabs: output.newTabs.map((tab: BrowserNewTab) => ({\n            targetId: tab.targetId,\n            ...(tab.url === undefined ? {} : { url: tab.url }),\n            ...(tab.title === undefined ? {} : { title: tab.title })\n          }))\n        }\n      : {})\n  };\n}\n\nfunction text(value: unknown): TextContent {\n  return {\n    type: \"text\",\n    text: typeof value === \"string\" ? value : JSON.stringify(value)\n  };\n}\n\n/**\n * The tool result pi stores and sends: the model projection of the run as\n * JSON (no `calls` log, bounded `logs`), and a screenshot as an image part\n * after it. pi-ai only sends the image to models that accept images.\n */\nfunction browserToolResult(\n  output: BrowserToolOutput\n): ToolExecutionResult<PiBrowserToolDetails> {\n  const isError = output.status === \"error\";\n  const screenshot = browserScreenshotOutput(output);\n  if (!screenshot) {\n    return {\n      output: [text(browserExecuteModelOutput(output).value)],\n      details: details(output),\n      ...(isError ? { isError } : {})\n    };\n  }\n\n  // No size check: a screenshot comes from one cdp.send result, which\n  // codemode caps at 1 MB, well under what model providers accept.\n  const bytes = Math.floor((screenshot.data.length * 3) / 4).toLocaleString();\n  // Keep the rest of the result (status, restarted, notice, newTabs) around\n  // the sentence that replaces the screenshot.\n  const summary = browserExecuteModelOutput({\n    ...output,\n    result: `Screenshot attached as an image (${screenshot.mediaType}, approximately ${bytes} bytes). If you can't see it, read the page with Runtime.evaluate instead.`\n  }).value;\n  const image: ImageContent = {\n    type: \"image\",\n    data: screenshot.data,\n    mimeType: screenshot.mediaType\n  };\n  return {\n    output: [text(summary), image],\n    details: details(output)\n  };\n}\n\nfunction conversationScope(api: ToolExecutionApi): string {\n  return `conversation:${api.conversationId}`;\n}\n\n/**\n * Create a pi-durable tool that lets the model drive a persistent browser\n * with JavaScript and the Chrome DevTools Protocol.\n *\n * Works like `browserTool` in `agents/browser/ai-sdk`: tabs, cookies, and\n * logins carry over between runs, `sessionId: \"active\"` addresses the tab the\n * model last worked in, and a replaced browser is reported as\n * `restarted: true`. A returned screenshot comes back as an image part that\n * the model sees, when its model accepts images.\n *\n * Each conversation gets its own active tab (see\n * {@link PiBrowserToolOptions.scope}). A conversation's calls run one at a\n * time, since they share its tab, and the tool doesn't rerun after an\n * eviction (its code may have clicked or\n * submitted something): pi gives the model an interrupted result instead.\n * Pass `ctx` unless the tool is built inside an Agent.\n *\n * @example\n * ```ts\n * import { Browser, browserRun } from \"agents/browser\";\n * import { browserTool } from \"agents/browser/pi\";\n *\n * export class MyAgent extends DurableObject<Env> {\n *   readonly browser = new Browser({ provider: browserRun(this.env.BROWSER) });\n *   readonly registry = createRegistry();\n *   readonly harness = new PiHarness({\n *     harness: ({ storage, context }) => {\n *       this.registry.install({\n *         name: \"browser\",\n *         tools: [\n *           browserTool({\n *             ctx: this.ctx,\n *             browser: this.browser,\n *             loader: this.env.LOADER\n *           })\n *         ]\n *       });\n *       return Harness.open(storage, { models, registry: this.registry }, context);\n *     }\n *   });\n *   readonly lifecycle = Lifecycle.install(this)\n *     .use(this.browser)\n *     .use(this.harness);\n * }\n * ```\n */\nexport function browserTool<TName extends string = \"browser\">(\n  options: PiBrowserToolOptions<TName>\n): PiBrowserTool<TName> {\n  const { scope = conversationScope, ...coreOptions } = options;\n  const core = createBrowserToolCore(coreOptions, {\n    screenshotHint:\n      \"To see a screenshot, return { type: 'browser_screenshot', mediaType, data } with data from Page.captureScreenshot and mediaType 'image/png', or 'image/jpeg' if you captured with format: 'jpeg'. The image is attached to the result.\"\n  });\n  return {\n    name: options.name ?? (\"browser\" as TName),\n    description: core.description,\n    parameters: browserToolParameters,\n    // A conversation's calls share its tab, so they must not interleave.\n    executionMode: \"sequential\",\n    // The code may have clicked or submitted something; don't run it twice.\n    replay: \"unsafe\",\n    async execute(args, api) {\n      const callScope = typeof scope === \"string\" ? scope : scope(api);\n      return browserToolResult(\n        await core.execute({ code: args.code }, callScope)\n      );\n    }\n  };\n}\n"],"mappings":";;;;AA2CA,MAAM,wBAAwB,KAAK,OAAO,EACxC,MAAM,KAAK,OAAO,EAChB,aAAa,6DACf,CAAC,EACH,CAAC;AAoBD,SAAS,QAAQ,QAAiD;CAChE,OAAO;EACL,aAAa,OAAO;EACpB,QAAQ,OAAO;EACf,GAAI,OAAO,YAAY,EAAE,WAAW,KAAc,IAAI,CAAC;EACvD,GAAI,OAAO,UACP,EACE,SAAS,OAAO,QAAQ,KAAK,SAAwB;GACnD,UAAU,IAAI;GACd,GAAI,IAAI,QAAQ,KAAA,IAAY,CAAC,IAAI,EAAE,KAAK,IAAI,IAAI;GAChD,GAAI,IAAI,UAAU,KAAA,IAAY,CAAC,IAAI,EAAE,OAAO,IAAI,MAAM;EACxD,EAAE,EACJ,IACA,CAAC;CACP;AACF;AAEA,SAAS,KAAK,OAA6B;CACzC,OAAO;EACL,MAAM;EACN,MAAM,OAAO,UAAU,WAAW,QAAQ,KAAK,UAAU,KAAK;CAChE;AACF;;;;;;AAOA,SAAS,kBACP,QAC2C;CAC3C,MAAM,UAAU,OAAO,WAAW;CAClC,MAAM,aAAa,wBAAwB,MAAM;CACjD,IAAI,CAAC,YACH,OAAO;EACL,QAAQ,CAAC,KAAK,0BAA0B,MAAM,CAAC,CAAC,KAAK,CAAC;EACtD,SAAS,QAAQ,MAAM;EACvB,GAAI,UAAU,EAAE,QAAQ,IAAI,CAAC;CAC/B;CAKF,MAAM,QAAQ,KAAK,MAAO,WAAW,KAAK,SAAS,IAAK,CAAC,CAAC,CAAC,eAAe;CAG1E,MAAM,UAAU,0BAA0B;EACxC,GAAG;EACH,QAAQ,oCAAoC,WAAW,UAAU,kBAAkB,MAAM;CAC3F,CAAC,CAAC,CAAC;CACH,MAAM,QAAsB;EAC1B,MAAM;EACN,MAAM,WAAW;EACjB,UAAU,WAAW;CACvB;CACA,OAAO;EACL,QAAQ,CAAC,KAAK,OAAO,GAAG,KAAK;EAC7B,SAAS,QAAQ,MAAM;CACzB;AACF;AAEA,SAAS,kBAAkB,KAA+B;CACxD,OAAO,gBAAgB,IAAI;AAC7B;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAgDA,SAAgB,YACd,SACsB;CACtB,MAAM,EAAE,QAAQ,mBAAmB,GAAG,gBAAgB;CACtD,MAAM,OAAO,sBAAsB,aAAa,EAC9C,gBACE,yOACJ,CAAC;CACD,OAAO;EACL,MAAM,QAAQ,QAAS;EACvB,aAAa,KAAK;EAClB,YAAY;EAEZ,eAAe;EAEf,QAAQ;EACR,MAAM,QAAQ,MAAM,KAAK;GACvB,MAAM,YAAY,OAAO,UAAU,WAAW,QAAQ,MAAM,GAAG;GAC/D,OAAO,kBACL,MAAM,KAAK,QAAQ,EAAE,MAAM,KAAK,KAAK,GAAG,SAAS,CACnD;EACF;CACF;AACF"}
+{"version":3,"file":"pi.js","names":[],"sources":["../../../src/browser/tools/pi.ts"],"sourcesContent":["import {\n  Type,\n  type ImageContent,\n  type TextContent\n} from \"@earendil-works/pi-ai\";\nimport type {\n  ToolExecutionApi,\n  ToolExecutionResult,\n  ToolRegistration\n} from \"@earendil-works/pi-durable\";\nimport {\n  createBrowserToolCore,\n  type BrowserToolOptions,\n  type BrowserToolOutput\n} from \"../browser-tool\";\nimport type { BrowserNewTab } from \"../session-connector\";\nimport {\n  browserExecuteModelOutput,\n  browserScreenshotOutput\n} from \"../tool-helpers\";\n\nexport type {\n  BrowserToolInput,\n  BrowserToolOptions,\n  BrowserToolOutput\n} from \"../browser-tool\";\nexport type { BrowserNewTab, BrowserSource } from \"../session-connector\";\n\nexport interface PiBrowserToolOptions<\n  TName extends string = \"browser\"\n> extends Omit<BrowserToolOptions, \"scope\"> {\n  /** The tool's name. Default `\"browser\"`. */\n  name?: TName;\n  /**\n   * Which calls share an active tab: a string for every call, or a function\n   * of the call. Defaults to `conversation:<id>` for the calling\n   * conversation, so each\n   * conversation (and each fork or subagent) on the Durable Object keeps its\n   * own tab, while cookies and logins stay shared.\n   */\n  scope?: string | ((api: ToolExecutionApi) => string);\n}\n\nconst browserToolParameters = Type.Object({\n  code: Type.String({\n    description: \"An async arrow function that drives the browser with `cdp`\"\n  })\n});\n\n/**\n * What a UI can show about a run without reading the content: the run's\n * outcome and what happened to the browser.\n */\nexport type PiBrowserToolDetails = {\n  executionId: string;\n  status: BrowserToolOutput[\"status\"];\n  restarted?: true;\n  newTabs?: { targetId: string; url?: string; title?: string }[];\n};\n\nexport type PiBrowserTool<TName extends string = \"browser\"> = ToolRegistration<\n  typeof browserToolParameters,\n  PiBrowserToolDetails\n> & {\n  readonly name: TName;\n};\n\nfunction details(output: BrowserToolOutput): PiBrowserToolDetails {\n  return {\n    executionId: output.executionId,\n    status: output.status,\n    ...(output.restarted ? { restarted: true as const } : {}),\n    ...(output.newTabs\n      ? {\n          newTabs: output.newTabs.map((tab: BrowserNewTab) => ({\n            targetId: tab.targetId,\n            ...(tab.url === undefined ? {} : { url: tab.url }),\n            ...(tab.title === undefined ? {} : { title: tab.title })\n          }))\n        }\n      : {})\n  };\n}\n\nfunction text(value: unknown): TextContent {\n  return {\n    type: \"text\",\n    text: typeof value === \"string\" ? value : JSON.stringify(value)\n  };\n}\n\n/**\n * The tool result pi stores and sends: the model projection of the run as\n * JSON (no `calls` log, bounded `logs`), and a screenshot as an image part\n * after it. pi-ai only sends the image to models that accept images.\n */\nfunction browserToolResult(\n  output: BrowserToolOutput\n): ToolExecutionResult<PiBrowserToolDetails> {\n  const isError = output.status === \"error\";\n  const screenshot = browserScreenshotOutput(output);\n  if (!screenshot) {\n    return {\n      output: [text(browserExecuteModelOutput(output).value)],\n      details: details(output),\n      ...(isError ? { isError } : {})\n    };\n  }\n\n  // No size check: a screenshot comes from one cdp.send result, which\n  // codemode caps at 1 MB, well under what model providers accept.\n  const bytes = Math.floor((screenshot.data.length * 3) / 4).toLocaleString();\n  // Keep the rest of the result (status, restarted, notice, newTabs) around\n  // the sentence that replaces the screenshot.\n  const summary = browserExecuteModelOutput({\n    ...output,\n    result: `Screenshot attached as an image (${screenshot.mediaType}, approximately ${bytes} bytes). If you can't see it, read the page with Runtime.evaluate instead.`\n  }).value;\n  const image: ImageContent = {\n    type: \"image\",\n    data: screenshot.data,\n    mimeType: screenshot.mediaType\n  };\n  return {\n    output: [text(summary), image],\n    details: details(output)\n  };\n}\n\nfunction conversationScope(api: ToolExecutionApi): string {\n  return `conversation:${api.conversationId}`;\n}\n\n/**\n * Create a pi-durable tool that lets the model drive a persistent browser\n * with JavaScript and the Chrome DevTools Protocol.\n *\n * Works like `browserTool` in `agents/browser/ai-sdk`: tabs, cookies, and\n * logins carry over between runs, `sessionId: \"active\"` addresses the tab the\n * model last worked in, and a replaced browser is reported as\n * `restarted: true`. A returned screenshot comes back as an image part that\n * the model sees, when its model accepts images.\n *\n * Each conversation gets its own active tab (see\n * {@link PiBrowserToolOptions.scope}). A conversation's calls run one at a\n * time, since they share its tab, and the tool doesn't rerun after an\n * eviction (its code may have clicked or\n * submitted something): pi gives the model an interrupted result instead.\n * Pass `ctx` unless the tool is built inside an Agent.\n *\n * @example\n * ```ts\n * import { Browser, browserRun } from \"agents/browser\";\n * import { browserTool } from \"agents/browser/pi\";\n *\n * export class MyAgent extends DurableObject<Env> {\n *   readonly browser = new Browser({ provider: browserRun(this.env.BROWSER) });\n *   readonly registry = createRegistry();\n *   readonly harness = new PiHarness({\n *     harness: ({ storage, context, onWake }) => {\n *       this.registry.install({\n *         name: \"browser\",\n *         tools: [\n *           browserTool({\n *             ctx: this.ctx,\n *             browser: this.browser,\n *             loader: this.env.LOADER\n *           })\n *         ]\n *       });\n *       return Harness.open(\n *         storage,\n *         { models, registry: this.registry, onWake },\n *         context\n *       );\n *     }\n *   });\n *   readonly lifecycle = Lifecycle.install(this)\n *     .use(this.browser)\n *     .use(this.harness);\n * }\n * ```\n */\nexport function browserTool<TName extends string = \"browser\">(\n  options: PiBrowserToolOptions<TName>\n): PiBrowserTool<TName> {\n  const { scope = conversationScope, ...coreOptions } = options;\n  const core = createBrowserToolCore(coreOptions, {\n    screenshotHint:\n      \"To see a screenshot, return { type: 'browser_screenshot', mediaType, data } with data from Page.captureScreenshot and mediaType 'image/png', or 'image/jpeg' if you captured with format: 'jpeg'. The image is attached to the result.\"\n  });\n  return {\n    name: options.name ?? (\"browser\" as TName),\n    description: core.description,\n    parameters: browserToolParameters,\n    // A conversation's calls share its tab, so they must not interleave.\n    executionMode: \"sequential\",\n    // The code may have clicked or submitted something; don't run it twice.\n    replay: \"unsafe\",\n    async execute(args, api) {\n      const callScope = typeof scope === \"string\" ? scope : scope(api);\n      return browserToolResult(\n        await core.execute({ code: args.code }, callScope)\n      );\n    }\n  };\n}\n"],"mappings":";;;;AA2CA,MAAM,wBAAwB,KAAK,OAAO,EACxC,MAAM,KAAK,OAAO,EAChB,aAAa,6DACf,CAAC,EACH,CAAC;AAoBD,SAAS,QAAQ,QAAiD;CAChE,OAAO;EACL,aAAa,OAAO;EACpB,QAAQ,OAAO;EACf,GAAI,OAAO,YAAY,EAAE,WAAW,KAAc,IAAI,CAAC;EACvD,GAAI,OAAO,UACP,EACE,SAAS,OAAO,QAAQ,KAAK,SAAwB;GACnD,UAAU,IAAI;GACd,GAAI,IAAI,QAAQ,KAAA,IAAY,CAAC,IAAI,EAAE,KAAK,IAAI,IAAI;GAChD,GAAI,IAAI,UAAU,KAAA,IAAY,CAAC,IAAI,EAAE,OAAO,IAAI,MAAM;EACxD,EAAE,EACJ,IACA,CAAC;CACP;AACF;AAEA,SAAS,KAAK,OAA6B;CACzC,OAAO;EACL,MAAM;EACN,MAAM,OAAO,UAAU,WAAW,QAAQ,KAAK,UAAU,KAAK;CAChE;AACF;;;;;;AAOA,SAAS,kBACP,QAC2C;CAC3C,MAAM,UAAU,OAAO,WAAW;CAClC,MAAM,aAAa,wBAAwB,MAAM;CACjD,IAAI,CAAC,YACH,OAAO;EACL,QAAQ,CAAC,KAAK,0BAA0B,MAAM,CAAC,CAAC,KAAK,CAAC;EACtD,SAAS,QAAQ,MAAM;EACvB,GAAI,UAAU,EAAE,QAAQ,IAAI,CAAC;CAC/B;CAKF,MAAM,QAAQ,KAAK,MAAO,WAAW,KAAK,SAAS,IAAK,CAAC,CAAC,CAAC,eAAe;CAG1E,MAAM,UAAU,0BAA0B;EACxC,GAAG;EACH,QAAQ,oCAAoC,WAAW,UAAU,kBAAkB,MAAM;CAC3F,CAAC,CAAC,CAAC;CACH,MAAM,QAAsB;EAC1B,MAAM;EACN,MAAM,WAAW;EACjB,UAAU,WAAW;CACvB;CACA,OAAO;EACL,QAAQ,CAAC,KAAK,OAAO,GAAG,KAAK;EAC7B,SAAS,QAAQ,MAAM;CACzB;AACF;AAEA,SAAS,kBAAkB,KAA+B;CACxD,OAAO,gBAAgB,IAAI;AAC7B;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAoDA,SAAgB,YACd,SACsB;CACtB,MAAM,EAAE,QAAQ,mBAAmB,GAAG,gBAAgB;CACtD,MAAM,OAAO,sBAAsB,aAAa,EAC9C,gBACE,yOACJ,CAAC;CACD,OAAO;EACL,MAAM,QAAQ,QAAS;EACvB,aAAa,KAAK;EAClB,YAAY;EAEZ,eAAe;EAEf,QAAQ;EACR,MAAM,QAAQ,MAAM,KAAK;GACvB,MAAM,YAAY,OAAO,UAAU,WAAW,QAAQ,MAAM,GAAG;GAC/D,OAAO,kBACL,MAAM,KAAK,QAAQ,EAAE,MAAM,KAAK,KAAK,GAAG,SAAS,CACnD;EACF;CACF;AACF"}

package/dist/chat-sdk/index.js

@@ -1,4 +1,4 @@
-import { i as getCurrentAgent, t as Agent } from "../src-6lr7MtM_.js";
+import { i as getCurrentAgent, t as Agent } from "../src-DfLzdwlK.js";
 //#region src/chat-sdk/agent.ts
 const NEXT_CLEANUP_AT_KEY = "next_cleanup_at";
 const CLEANUP_SCHEDULE_ID_KEY = "cleanup_schedule_id";