Add MCP tool to list all available tasks and enable fuzzy task matching

This enables LLMs to figure out which task you are referring to, without needing to specify the full task ID

Author: myftija

packages/cli-v3/src/dev/mcpServer.ts

@@ -1,11 +1,13 @@
+import polka from "polka";
 import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
 import { SSEServerTransport } from "@modelcontextprotocol/sdk/server/sse.js";
-import { optional, union, z } from "zod";
+import { z } from "zod";
 import { logger } from "../utilities/logger.js";
 import { CliApiClient } from "../apiClient.js";
 import { ApiClient, RunStatus } from "@trigger.dev/core/v3";
-import polka from "polka";
+import { eventBus } from "../utilities/eventBus.js";
 
+let allTaskIds: string[] = [];
 let projectRef: string;
 let dashboardUrl: string;
 // there is some overlap between `ApiClient` and `CliApiClient` which is not ideal
@@ -20,6 +22,22 @@ const server = new McpServer({
   version: "1.0.0",
 });
 
+// The `list-all-tasks` tool primarily helps to enable fuzzy matching of task IDs (names).
+// This way, one doesn't need to specify the full task ID and rather let the LLM figure it out.
+// This could be a good fit for the `resource` entity in MCP.
+// Also, a custom `prompt` entity could be useful to instruct the LLM to prompt the user
+// for selecting a task from a list of matching tasks, when the confidence for an exact match is low.
+server.tool("list-all-tasks", "List all available task IDs in the worker.", async () => {
+  return {
+    content: [
+      {
+        text: JSON.stringify(allTaskIds, null, 2),
+        type: "text",
+      },
+    ],
+  };
+});
+
 server.tool(
   "trigger-task",
   "Trigger a task",
@@ -213,6 +231,10 @@ app.post("/messages", (req, res) => {
   }
 });
 
+eventBus.on("backgroundWorkerInitialized", (worker) => {
+  allTaskIds = worker.manifest?.tasks.map((task) => task.id) ?? [];
+});
+
 export const startMcpServer = async (options: {
   port: number;
   cliApiClient: CliApiClient;