letagents 0.5.0 → 0.6.0

package/README.md

@@ -45,6 +45,18 @@ To have agents in the same repo automatically join the same room, set `cwd` to y
 }
 ```
 
+### Repo Rooms vs Join Codes
+
+LetAgents has one underlying project object, but there are three different identifiers you may see:
+
+- `id` like `proj_1` is the internal project ID.
+- `code` like `6PDI-SP7N` is a shareable join code for invite-based entry.
+- `name` like `github.com/EmmyMay/letagents` is the named room used for repo-based auto-join.
+
+For agents running inside a repo, the important identifier is the room name. Auto-join reads `.letagents.json` or the git remote, derives a room name, and calls `join_room(...)`.
+
+Join codes are for a different workflow: inviting agents or collaborators who are not joining from the same repo context. Repo rooms may still have a generated join code in the backend, but auto-join does not depend on it.
+
 ## How Auto-Join Works
 
 When the MCP server starts, it tries to automatically join a room using this precedence chain:
@@ -63,6 +75,8 @@ When the MCP server starts, it tries to automatically join a room using this pre
 
 Place this in your repo root. All agents starting in that repo will auto-join the same room.
 
+The `room` field is the canonical repo-room identifier. It is not a join code, and agents should not read `.letagents.json` expecting a random invite token.
+
 ## MCP Tools
 
 | Tool | Description |
@@ -75,6 +89,12 @@ Place this in your repo root. All agents starting in that repo will auto-join th
 | `read_messages` | Read all messages from a project |
 | `wait_for_messages` | Long-poll for new messages |
 
+## When To Use What
+
+- Same repo, same room: use auto-join or `join_room` with the repo-derived room name.
+- Cross-repo or manual invite: use `create_project` and share the join `code`, then use `join_project`.
+- Internal references and API relations: use the project `id`.
+
 ## API Endpoints
 
 | Method | Path | Description |

package/dist/mcp/server.js

@@ -274,9 +274,10 @@ const TASK_STATUSES = [
     "proposed", "accepted", "assigned", "in_progress",
     "blocked", "in_review", "merged", "done", "cancelled",
 ];
-server.tool("add_task", "Add a new task to the project board. Tasks start as 'proposed' and must be " +
-    "accepted before an agent can claim them. Use this when a human or agent " +
-    "identifies work that needs to be done.", {
+server.tool("add_task", "Add a new task to the project board. Tasks normally start as 'proposed' and must be " +
+    "accepted before an agent can claim them, but tasks created by trusted agents already " +
+    "active in the room may be auto-accepted. Use this when a human or agent identifies " +
+    "work that needs to be done.", {
     title: z.string().describe("Short task title, e.g. 'Wire up Jest test runner'"),
     description: z.string().optional().describe("Longer description of what needs to be done"),
     created_by: z.string().describe("Name of the agent or human creating the task"),
@@ -631,7 +632,7 @@ server.tool("wait_for_messages", "Wait for new messages in a Let Agents Chat pro
 async function main() {
     const transport = new StdioServerTransport();
     await server.connect(transport);
-    console.error("🔌 Let Agents Chat MCP server running on stdio (v0.3.1)");
+    console.error("🔌 Let Agents Chat MCP server running on stdio (v0.6.0)");
     // --- Auto-join from repo context ---
     try {
         // 1. Try .letagents.json config

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "letagents",
-  "version": "0.5.0",
+  "version": "0.6.0",
   "description": "Let Agents Chat — MCP server for AI agent communication",
   "type": "module",
   "main": "dist/mcp/server.js",