Get the FREE Ultimate OpenClaw Setup Guide →

ACP

A bridge server that connects Agent Communication Protocol (ACP) agents with Model Context Protocol (MCP) clients, enabling seamless integration between ACP-based AI agents and MCP-compatible tools like Claude Desktop.

Installation
Run this command in your terminal to add the MCP server to Claude Code.
Run in terminal:
Command
claude mcp add --transport stdio gongrzhe-acp-mcp-server uvx acp-mcp-server

How to use

ACP-MCP-Server acts as a bridge between ACP agents and MCP clients. It exposes tools that allow you to discover ACP agents, execute them, and route requests intelligently to the right agent. You can run the server in several transport modes: STDIO (the default, suitable for Claude Desktop), Server-Sent Events (SSE) for streaming web clients, or the streamable HTTP transport for REST-like interactions. The project also includes command-line options to connect to a different ACP server and to configure host/port for HTTP-based transports. When integrated with Claude Desktop, you add a tiny configuration snippet that points to the bridge so Claude can route MCP requests to ACP agents.

Key capabilities include agent discovery and management (discover_acp_agents, get_agent_info), agent execution (run_acp_agent, get_async_run_result, list_active_runs), smart routing (smart_route_request, add_routing_rule, list_routing_strategies), interactive sessions (start_interactive_agent, provide_user_input, list_pending_interactions), and message processing utilities (convert_acp_message, analyze_message_content). The bridge handles both synchronous and asynchronous operations, supports multi-turn interactive sessions, and can handle text, images, and other content types via its multi-modal design.

How to install

Prerequisites:

  • Python 3.11+ installed on your system
  • Pip (Python package manager)
  • Optional: uvx for isolated execution of Python packages

Install from PyPI:

pip install acp-mcp-server

Or run with uvx (isolated execution):

uvx acp-mcp-server

If you prefer to run from source, clone the repo (or download the package) and install dependencies in a virtual environment:

python3 -m venv venv
source venv/bin/activate  # on Windows use .\venv\Scripts\activate
pip install -r requirements.txt  # if a requirements file is provided
pip install acp-mcp-server

Usage after installation (examples):

# Run stdio (default transport)
acp-mcp-server

# Run with SSE transport
acp-mcp-server --transport sse --port 8000

# Run with HTTP transport (streamable HTTP)
acp-mcp-server --transport streamable-http --host 0.0.0.0 --port 9000

# Connect to a different ACP server
acp-mcp-server --acp-url http://localhost:8001

If you are using uvx for isolated execution, the example configuration snippet in Claude Desktop can be used as shown in the README:

{
  "mcpServers": {
    "acp-bridge": {
      "command": "uvx",
      "args": ["acp-mcp-server"]
    }
  }
}

Additional notes

Environment variable ACP_BASE_URL controls the ACP server endpoint the bridge connects to (default http://localhost:8000). If you run the server behind a proxy or inside Kubernetes, ensure port mappings and host settings align with your deployment. The Docker instructions in the README provide multi-container setups via docker-compose for HTTP and SSE variants, handy for production or dev workflows. When troubleshooting, verify that the ACP server is reachable at ACP_BASE_URL and that the container or environment has network access to that endpoint. If you encounter transport-related issues, check that the chosen transport (stdio, sse, or streamable-http) is properly configured and that the target port is open and not blocked by a firewall.

Related MCP Servers

Sponsor this space

Reach thousands of developers