@launchchair/agent 0.5.0 → 0.5.1

package/AGENTS.md

@@ -36,6 +36,8 @@ curl -fsSL "https://www.launchchair.io/api/agent/v1/bridge/local-bridge.mjs" -o
 curl -fsSL "https://www.launchchair.io/api/agent/v1/bridge/launchchair-cli.mjs" -o "$HOME/launchchair-agent/launchchair-cli.mjs"
 curl -fsSL "https://www.launchchair.io/api/agent/v1/bridge/mcp-bridge-server.mjs" -o "$HOME/launchchair-agent/mcp-bridge-server.mjs"
 curl -fsSL "https://www.launchchair.io/api/agent/v1/bridge/mcp-capabilities.mjs" -o "$HOME/launchchair-agent/mcp-capabilities.mjs"
+curl -fsSL "https://www.launchchair.io/api/agent/v1/bridge/mcp-widgets.mjs" -o "$HOME/launchchair-agent/mcp-widgets.mjs"
+curl -fsSL "https://www.launchchair.io/api/agent/v1/bridge/mcp-widget.html" -o "$HOME/launchchair-agent/mcp-widget.html"
 curl -fsSL "https://www.launchchair.io/api/agent/v1/bridge/project-workspace.mjs" -o "$HOME/launchchair-agent/project-workspace.mjs"
 curl -fsSL "https://www.launchchair.io/api/agent/v1/bridge/runner-command.mjs" -o "$HOME/launchchair-agent/runner-command.mjs"
 curl -fsSL "https://www.launchchair.io/api/agent/v1/bridge/codex-runner.mjs" -o "$HOME/launchchair-agent/codex-runner.mjs"
@@ -131,6 +133,7 @@ Prefer `launchchair setup codex` so the CLI writes native config and verifies MC
 6. Use `launchchair_run_once` or `executionMode=local_runner` only when the user explicitly requests a separate local runner.
 7. If LaunchChair asks for a user-choice bucket such as wedge, pricing, or feature direction, call `launchchair_select_choice`. Do not use generic SPEC_UPDATE apply for user-choice buckets.
 8. When the user asks for validation scores or wants to compare ideas, call `launchchair_get_scorecards`. Report validation scorecards separately from generated direction-option scores.
+9. In hosts that support MCP Apps, present `launchchair_show_choices` at a choice step and after a selection, `launchchair_show_phase_synthesis` after phase completion, and `launchchair_show_validation_scores` for validation results. Their Next action controls return work to the host agent. If UI is unsupported, summarize the structured result and show the returned URLs.
 
 ## Consent Behavior
 

package/README.md

@@ -12,12 +12,30 @@ launchchair --version
 launchchair doctor
 ```
 
-Version 0.5.0 includes the expanded MCP, browser login, private token management, host-agent execution, idea preservation, and optional validation gates. Existing installations must update the package and restart their MCP connection to load the new tools.
+Version 0.5.1 adds tabbed MCP Apps widgets for saved choices, rich phase synthesis, and Product/Feature validation scorecards to the expanded MCP, browser login, private token management, host-agent execution, idea preservation, and optional validation gates. Existing installations must update the package and restart their MCP connection to load the new tools.
 
 After updating, run `launchchair setup codex --skip-login` or `launchchair setup claude --skip-login` to refresh the installed path and verify tool discovery. Then restart/reload that client. Your saved account login and project bindings remain in `~/.launchchair/config.json`. Older tokens still work within their existing scopes; run `launchchair login` again with workspace-owner approval if you need the new `tokens.manage` scope.
 
 For local release testing, run `npm pack ./packages/launchchair-agent --pack-destination /tmp` from the repository root and install the resulting tarball into an isolated npm prefix before publishing.
 
+## Agent response widgets
+
+The bridge supports LaunchChair AUX through MCP Apps. Three read-only tools return tabbed widgets using the app's Chakra theme, shared choice cards, and existing strategic scorecards:
+
+- `launchchair_show_choices` (`projectId`, `phaseKey`, `bucketKey`): generated options and the saved choice.
+- `launchchair_show_phase_synthesis` (`projectId`, `phaseKey`): the app's phase-specific synthesis cards, native tabs and charts, with verified completion state.
+- `launchchair_show_validation_scores` (`projectId`): validation scorecards, evidence inputs, and separate direction scores for Product and Feature projects.
+
+Every widget includes **Next action** controls. Saving a choice calls the existing `launchchair_select_choice` tool, then rereads the saved selection. Continue/review actions send a message to the host agent; the host executes queued prompts with its tools. The iframe never receives a token or calls the API directly. Existing scopes, entitlements, idea policies, validation gates, and downstream resets still apply.
+
+The content pane scrolls at 560px on desktop and 440px on smaller screens using the app's scrollbar styles. The next-action footer stays outside that pane. Selected choices reuse the same choice cards, and validation thresholds live inside the scorecard footer. Synthesis reads the sanitized spec through the existing export permission; competitor editing and workspace notes remain in the authenticated app.
+
+Rendering uses `_meta.ui.resourceUri`, `text/html;profile=mcp-app` resources, and the MCP Apps JSON-RPC host bridge. Compatible ChatGPT hosts can also use the feature-detected `window.openai` aliases. Text-only clients receive the full structured result and URLs. Static templates can be discovered while logged out; project reads still require authentication.
+
+From the repository root, run `npm run agent:widgets:build` after changing widget components or the theme. This rebuilds the self-contained HTML in both bridge distributions. `npm run agent:widgets:preview -- /absolute/path/examples.html` builds interactive local examples with labeled sample data. Nothing is published by these commands.
+
+The choice widget requires the bucket API's additive `choice` fields. Synthesis requests the existing export endpoint with `?format=json`; default Markdown downloads are unchanged. Test against a matching local API before releasing. Release the matching API changes before publishing the bridge package so installed clients can read the new choice and synthesis payloads.
+
 ## Login
 
 Approve the CLI once in your browser. LaunchChair creates a scoped Agent API token after approval and the CLI stores it locally:

package/launchchair-cli.mjs

@@ -20,7 +20,7 @@ import {
   unbindProjectWorkspace,
 } from "./project-workspace.mjs";
 
-const VERSION = "0.5.0";
+const VERSION = "0.5.1";
 const SCRIPT_DIR = path.dirname(fileURLToPath(import.meta.url));
 const DEFAULT_BASE_URL = "https://www.launchchair.io";
 const CONFIG_DIR = process.env.LAUNCHCHAIR_CONFIG_DIR?.trim()

package/local-bridge.mjs

@@ -18,7 +18,7 @@ import { fileURLToPath } from "node:url";
 import { resolveProjectWorkspace } from "./project-workspace.mjs";
 import { buildRunnerCommand, parseRunnerCommand } from "./runner-command.mjs";
 
-const VERSION = "0.5.0";
+const VERSION = "0.5.1";
 const CLIENT_TYPES = ["codex", "chatgpt", "claude", "claude_code", "other"];
 const AGENT_SCRIPT_DIR = path.dirname(fileURLToPath(import.meta.url));
 const CONFIG_DIR = process.env.LAUNCHCHAIR_CONFIG_DIR?.trim()

package/mcp-bridge-server.mjs

@@ -20,9 +20,11 @@ import { fileURLToPath } from "node:url";
 import { resolveProjectWorkspace } from "./project-workspace.mjs";
 import { buildRunnerCommand, parseRunnerCommand } from "./runner-command.mjs";
 import { createMcpCapabilities, EXTRA_MCP_TOOLS, MCP_WORKFLOW_INSTRUCTIONS, VALIDATION_GATE_SCHEMA } from "./mcp-capabilities.mjs";
+import { MCP_WIDGET_TOOLS, MCP_WIDGET_INSTRUCTIONS, listWidgetResources, readWidgetResource, callWidgetTool } from "./mcp-widgets.mjs";
 
-const VERSION = "0.5.0";
+const VERSION = "0.5.1";
 const SERVER_NAME = "launchchair-agent-mcp";
+const SUPPORTED_PROTOCOL_VERSIONS = new Set(["2025-11-25", "2025-06-18", "2025-03-26", "2024-11-05"]);
 const AGENT_SCRIPT_DIR = path.dirname(fileURLToPath(import.meta.url));
 const CONFIG_DIR = process.env.LAUNCHCHAIR_CONFIG_DIR?.trim()
   ? path.resolve(process.env.LAUNCHCHAIR_CONFIG_DIR.trim())
@@ -181,7 +183,7 @@ function loadStatus() {
 
 function serverInstructions(status) {
   const featureInstructions = String(status?.capabilities?.mcpServerInstructions ?? "").trim();
-  return `${BASE_SERVER_INSTRUCTIONS} ${featureInstructions} ${MCP_WORKFLOW_INSTRUCTIONS}`;
+  return `${BASE_SERVER_INSTRUCTIONS} ${featureInstructions} ${MCP_WORKFLOW_INSTRUCTIONS} ${MCP_WIDGET_INSTRUCTIONS}`;
 }
 
 function targetForRun(run) {
@@ -989,6 +991,7 @@ const READ_ONLY_TOOLS = new Set(["launchchair_status", "launchchair_list_project
 function annotateTool(tool) {
   return {
     ...tool,
+    ...(tool.name === "launchchair_select_choice" ? { _meta: { ...tool._meta, ui: { visibility: ["model", "app"] }, "openai/widgetAccessible": true } } : {}),
     annotations: {
       ...(tool.annotations ?? {}),
       readOnlyHint: tool.annotations?.readOnlyHint ?? READ_ONLY_TOOLS.has(tool.name),
@@ -1034,6 +1037,7 @@ function writeError(id, code, message, data = undefined) {
 
 function textResult(payload) {
   return {
+    ...(payload && typeof payload === "object" && !Array.isArray(payload) ? { structuredContent: payload } : {}),
     content: [
       {
         type: "text",
@@ -1054,9 +1058,10 @@ async function handleRpc(request) {
   if (method === "initialize") {
     const status = await loadStatus();
     writeResult(id, {
-      protocolVersion: "2024-11-05",
+      protocolVersion: SUPPORTED_PROTOCOL_VERSIONS.has(request?.params?.protocolVersion) ? request.params.protocolVersion : "2025-11-25",
       capabilities: {
         tools: { listChanged: true },
+        resources: {},
       },
       serverInfo: {
         name: SERVER_NAME,
@@ -1072,12 +1077,28 @@ async function handleRpc(request) {
     return;
   }
 
+  if (method === "resources/list") {
+    writeResult(id, { resources: listWidgetResources() });
+    return;
+  }
+
+  if (method === "resources/read") {
+    try { writeResult(id, readWidgetResource(request?.params?.uri)); }
+    catch (error) { writeError(id, -32002, error.message); }
+    return;
+  }
+
+  if (method === "ping") {
+    writeResult(id, {});
+    return;
+  }
+
   if (method === "tools/list") {
     const status = await loadStatus();
     const overrides = Array.isArray(status?.capabilities?.mcpToolOverrides)
       ? status.capabilities.mcpToolOverrides
       : [];
-    const tools = mergeToolOverrides([...MCP_TOOLS, ...EXTRA_MCP_TOOLS], overrides);
+    const tools = [...mergeToolOverrides([...MCP_TOOLS, ...EXTRA_MCP_TOOLS], overrides), ...MCP_WIDGET_TOOLS];
     writeResult(id, { tools });
     return;
   }
@@ -1087,6 +1108,10 @@ async function handleRpc(request) {
     const toolName = String(params.name ?? "");
     const args = typeof params.arguments === "object" && params.arguments ? params.arguments : {};
     try {
+      if (MCP_WIDGET_TOOLS.some((tool) => tool.name === toolName)) {
+        writeResult(id, await callWidgetTool(toolName, args, { api, resolveProjectId, baseUrl: config.baseUrl }));
+        return;
+      }
       if (EXTRA_MCP_TOOLS.some((tool) => tool.name === toolName)) {
         writeResult(id, textResult(await capabilities.dispatch(toolName, args)));
         return;