Package Exports

cursor-claude-messenger
cursor-claude-messenger/dist/index.js

This package does not declare an exports field, so the exports above have been automatically detected and optimized by JSPM instead. If any package subpath is missing, it is recommended to post an issue to the original package (cursor-claude-messenger) to support the "exports" field. If that is not possible, create a JSPM override to customize the exports field for this package.

Readme

agent-messenger

MCP server for AI agent-to-agent messaging. Lets Cursor and Claude Code agents send messages, reply in threads, and coordinate work — backed by Beads for persistent, version-controlled storage.

Quick Start

Prerequisites: Node.js 20+, Beads (bd CLI) v1.0.0+, Dolt v1.85.0+

# Install globally (the CLI command is "agent-messenger")
npm install -g cursor-claude-messenger

# Set up in your project
cd your-project
agent-messenger init

# Restart Cursor. Done.

Windows: If agent-messenger isn't recognized after install, the npm global bin directory likely isn't in your PATH. Run npm prefix -g to find it (usually C:\Users\<you>\AppData\Roaming\npm), then add it permanently:
[Environment]::SetEnvironmentVariable("PATH", $env:PATH + ";$(npm prefix -g)", "User")
Restart your terminal after.

The init command handles Beads initialization, MCP config generation, Cursor rules, and Claude Code skills. If something doesn't work, run:

agent-messenger doctor

How It Works

Both agents connect to the same MCP server with different identities. Messages route via labels (to:, from:, unread) stored in a Beads (Dolt) database. Threading uses replies_to graph links. Channels isolate conversations when multiple agent pairs are active.

┌──────────────┐         ┌─────────────────────┐         ┌──────────────┐
│    Cursor     │◄─stdio─►│  agent-messenger MCP │◄─stdio─►│  Claude Code  │
│  cursor-opus  │         │                     │         │  claude-code  │
│               │         │  send_message       │         │               │
│ #cm #sm #id   │         │  check_inbox        │         │ /cm /sm /id   │
└──────────────┘         │  reply / get_thread  │         └──────────────┘
                          │  set_channel / ...   │
                          │                     │
                          │  ┌───────────────┐  │
                          │  │ Beads (bd CLI) │  │
                          │  │ .beads/ Dolt DB│  │
                          │  └───────────────┘  │
                          └─────────────────────┘

Tools (21)

Messaging:

Tool	Description
`send_message`	Send a message (supports `task_id` for linking)
`check_inbox`	Check for unread messages
`reply`	Reply to a message (auto-threads, optional `task_id`)
`get_thread`	Get full conversation thread
`list_conversations`	List all conversations
`mark_read`	Mark a message as read
`set_channel`	Join a channel for multi-agent isolation
`set_identity`	Rename this agent instance
`whoami`	Show identity, base ID, and channel

Tasks:

Tool	Description
`create_task`	Create a task in Beads
`create_epic`	Create an epic for phased planning
`list_tasks`	List tasks with filters (status, priority, ready-only)
`show_task`	Show task details and linked messages
`update_task`	Update status, notes, labels, priority, or assignee (smart-routes close/reopen)
`claim_task`	Atomically assign and start a task
`close_task`	Close a completed task
`manage_deps`	Add, remove, or list dependencies between tasks
`blocked_tasks`	Show tasks blocked by unresolved dependencies
`project_stats`	Project health snapshot: counts, ready work, lead time

Discovery:

Tool	Description
`list_agents`	Show agents currently online

Shortcuts

Cursor (via rules): #help #cm #sm #ch #id #wi #ct #lt #st #rt #la #orchestrate

Claude Code (via skills): /am /cm /sm /ch /id /wi /ct /lt /st /rt /la /orchestrate

Orchestrate Workflow

#orchestrate <feature> (Cursor) or /orchestrate (CC) activates a structured development workflow pairing Cursor as orchestrator (brainstormer, designer, spec writer, plan reviewer) and Claude Code as implementer (spec verifier, plan writer, coder).

Built on the superpowers process: Cursor uses brainstorming for design rigor, CC uses writing-plans and subagent-driven-development for implementation with TDD. Abandon at any step — no state to clean up.

See docs/setup-guide.md for the full workflow description.

Identity & Multi-Agent

Each agent gets a unique session ID on startup (e.g. claude-code-a3f2). The base ID (claude-code) is shared across all instances — messages to the base ID reach every instance. Use set_identity (#id / /id) to pick a memorable name like cc-design.

When multiple agent windows/terminals are open in the same project:

Name them: #id cursor-design, /id cc-design — then address by name

Use channels: #ch design-review, /ch design-review — only paired agents see messages

CLI Commands

Command	Description
`agent-messenger init`	Set up agent-messenger in current project
`agent-messenger doctor`	Diagnose common setup issues
`agent-messenger status`	Show unread counts, agents, and channels
`agent-messenger log`	View message history (filters, thread view)
`agent-messenger help`	Show help

Init options

Flag	Default	Description
`--cursor-id <id>`	`cursor-opus`	Cursor agent ID
`--cc-id <id>`	`claude-code`	Claude Code agent ID
`--dry-run`		Preview changes without writing
`--skip-beads`		Skip Beads/Dolt setup

Troubleshooting

Run agent-messenger doctor first — it checks everything automatically.

Problem	Fix
`agent-messenger` not recognized (Windows)	npm global bin isn't in PATH — see the Windows note above
MCP server disabled after PC sleep/wake	Toggle it off and back on in Cursor Settings > Tools & MCP (known Cursor issue)
"driver: bad connection"	Run `bd dolt start` — the Dolt server isn't running
MCP not appearing in Cursor	Restart Cursor; check `.cursor/mcp.json` exists with correct paths

See docs/setup-guide.md for the full troubleshooting table, manual setup, and detailed usage.

Contributing

See docs/setup-guide.md for development setup. The project uses Beads for issue tracking — run bd ready to find available work.

License

MIT