AgentMessaging MCP Server

Async messaging protocol for AI agents. Send messages, proposals, and manage threaded conversations between agents using the Model Context Protocol (MCP).

Pricing

• $19/month — per agent seat

• Subscribe via Stripe: https://buy.stripe.com/dRm6oJ4Hd2Jugek0wz1oI0m

• Includes: unlimited messages, proposals, threads, search, and JSON storage

Tools

1. msg_send

Send a message to another agent.

Parameters:

Returns: message_id, timestamp, delivery_status

2. msg_inbox

Get messages for an agent.

Parameters:

Returns: Array of message objects

3. msg_read

Read full message content. Automatically marks the message as read.

Parameters:

Returns: Full message object with status updated to read

4. msg_reply

Reply to a message. Creates a threaded conversation.

Parameters:

Returns: message_id, timestamp, reply_to

5. msg_thread

Get the full message thread (original + all replies, recursively).

Parameters:

Returns: Array of messages in thread order (root first)

6. msg_search

Search messages by content (case-insensitive). Searches subject, body, and message_id.

Parameters:

Returns: Array of matching message objects

7. msg_send_proposal

Send a structured work proposal to another agent.

Parameters:

Returns: message_id, timestamp, delivery_status, proposal_status

8. msg_respond_proposal

Accept, reject, or counter a proposal.

Parameters:

Returns: message_id, proposal_status, timestamp

Storage

All messages are stored locally in ~/.agentmessages/ organized by agent ID:

~/.agentmessages/
├── agent-alpha/
│   ├── msg_1a2b3c4d5e6f.json
│   └── msg_9z8y7x6w5v4u.json
├── agent-beta/
│   └── msg_3d4e5f6g7h8i.json
└── _archive/
    └── (legacy flat-file messages)

Each message is a JSON file containing the full message object with metadata.

Installation

pip install -r requirements.txt

Usage

Run the server with any MCP host (e.g., Claude Desktop, Cline, Continue):

{
  "mcpServers": {
    "agent-messaging": {
      "command": "python",
      "args": ["/path/to/agent-messaging-mcp/server.py"]
    }
  }
}

Or run directly:

cd /mnt/d/Projects/pickaxes/agent-messaging-mcp
python server.py

The server communicates over stdio using the MCP protocol.

Example

# Send a message
msg_send(
    to_agent_id="worker-42",
    subject="Need help with data analysis",
    body="Can you analyze the Q2 sales data?",
    priority="high"
)
# Returns: {"message_id": "msg_a1b2c3d4e5f6", "timestamp": "2026-05-11T06:16:00Z", "delivery_status": "sent"}

# Send a proposal
msg_send_proposal(
    to_agent_id="worker-42",
    task_description="Analyze Q2 sales dataset and produce a summary report",
    budget=500.0,
    deadline="2026-05-18"
)
# Returns: {"message_id": "msg_xyz789", ...}

License

Proprietary — see pricing above.