mcp-agentapi

# MCP Agent API A Model Context Protocol (MCP) server that bridges MCP clients with AI agents through the [Agent API](https://github.com/coder/agentapi) by [Coder](https://github.com/coder). ``` ┌─────────────────┐ ┌──────────────┐ ┌─────────────┐ │ MCP Client │ │ MCP Server │ │ Agent API │ │ (Windsurf, │◄───►│ (This tool) │◄───►│ (AI Agent) │ │ Augment, etc.) │ │ │ │ │ └─────────────────┘ └──────────────┘ └─────────────┘ ``` This package enables you to use AI agents like Goose, Aider, and Claude through any MCP-compatible client. It implements the [Model Context Protocol](https://github.com/modelcontextprotocol/protocol) specification to provide a standardized interface for controlling multiple AI agents. ## 🚀 Key Features - **Multi-Agent Support**: Control Goose, Aider, Claude, and other agents through a unified interface - **Agent Lifecycle Management**: Detect, install, start, stop, and restart agents - **Message Handling**: Seamlessly send and receive messages between clients and agents - **Health Monitoring**: Monitor agent and API health status - **Flexible Transport**: Support for both stdio and SSE transport protocols - **Command-Line Interface**: Comprehensive CLI for all operations ## 📦 Installation ```bash # Using uv (Recommended) pip install uv uv pip install mcp-agentapi # Using pip pip install mcp-agentapi # From source git clone https://github.com/lord-dubious/mcp-agentapi.git cd mcp-agentapi ./build.sh uv pip install dist/*.whl ``` For development installation: ```bash uv pip install -e ".[dev]" ``` ## 🔌 Client Configuration ### Quick Setup Add to your MCP client configuration (Claude Desktop, Windsurf, Augment, etc.): ```json { "mcpServers": { "agent-controller": { "command": "python", "args": ["-m", "mcp_agentapi", "--transport", "stdio"], "env": { "TRANSPORT": "stdio" } } } } ``` ### Agent-Specific Configuration…