Agent Communication

Provides room-based, Slack-like inter-agent messaging via MCP with room management, message exchange, long-polling, and admin tools.

typescript

GitHub Stars

typescript

Language

5 months ago

First Indexed

2 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "mkxultra-agent-communication-mcp": {
      "command": "npx",
      "args": [
        "agent-communication-mcp"
      ],
      "env": {
        "AGENT_COMM_DATA_DIR": "YOUR_DATA_DIR",
        "AGENT_COMM_MAX_ROOMS": "100",
        "AGENT_COMM_LOCK_TIMEOUT": "5000",
        "AGENT_COMM_MAX_MESSAGES": "10000",
        "AGENT_COMM_WAIT_TIMEOUT": "120000"
      }
    }
  }
}

You can run an MCP server that enables multiple AI agents to communicate in room-based channels, similar to Slack. It provides room management, message exchange, long-polling for new messages, and administrative controls to keep conversations consistent and organized.

How to use

Configure your MCP client to connect to the agent communication server using a local stdio instance. You will connect agents to rooms, send messages, fetch messages, wait for new messages, manage rooms, and monitor system status. Use the provided tools to list rooms, create rooms, enter or leave rooms, send messages with optional metadata, retrieve messages, wait for new messages, and perform maintenance tasks like clearing messages or checking status.

How to install

# Prerequisites
- Node.js (LTS) and npm installed

# 1) Install the MCP server package
npm install agent-communication-mcp

# 2) Build the TypeScript source (if you cloned the repo or are working from source)
npm install
npm run build

# Optional: run in development mode (if provided by your setup)
npm run dev

You can also run the MCP server from a local build by invoking the built entry point with Node. For a local install, use the same commands but point to the built distribution.

## Additional sections

Configuration and runtime details are shown below, including how to connect via stdio commands and which environment variables to provide for proper operation.

## Configuration and runtime details

The server can be run as a local process using standard I/O (stdio). Use one of the following runtime commands to start the MCP server locally.

## Troubleshooting

If you encounter file locking or timeout issues, adjust the corresponding environment variables provided below. Ensure rooms exist before attempting to enter them, and verify agents are in the correct rooms before sending messages.

## Available tools

### list\_rooms

Fetches the list of all rooms or only those that a specific agent has joined

### create\_room

Creates a new room with a given name and description

### enter\_room

Adds an agent to a room with a profile describing its role and capabilities

### leave\_room

Removes an agent from a room

### list\_room\_users

Returns the list of users currently in a specified room

### send\_message

Sends a message to a room with optional metadata like priority

### get\_messages

Retrieves messages from a room, with options to filter by agent or mentions

### wait\_for\_messages

Long-polling to wait for new messages in a room for a specific agent

### get\_status

Retrieves system-wide or per-room status information

### clear\_room\_messages

Clears messages in a specified room after confirmation

Built by

VeilStrat

AI signals for GTM teams