MCP

MCP (Model Context Protocol) is an open protocol that defines how AI Agents interact with external tools. XClaw natively supports MCP, which means you can connect virtually any external service through a unified interface.

Why does this matter? Because an Agent's capability ceiling depends on the tools it can reach. Without MCP, every tool needs custom integration code. With MCP, any service that implements the protocol can plug into the Agent's toolchain instantly.

XClaw supports two MCP transport modes: stdio (local process communication) and SSE (remote HTTP streaming). Most community MCP servers use stdio — typically a Node.js process you start with a single npx command.

A connected MCP server exposes a set of tools to the Agent. The Agent autonomously decides when to use which tool during reasoning. You don't need to manually tell it "use the GitHub tool to check this" — it judges from conversation context on its own.

Credential management is secure. API keys and tokens are encrypted and stored locally. Each MCP connection has independent status monitoring, and XClaw will automatically reconnect if a connection drops.

How to

Open Settings > MCP Connections and click "Add MCP Server." Choose the transport mode — most community MCP servers use stdio, so just enter the launch command (e.g., npx -y @modelcontextprotocol/server-github). For remote services, pick SSE mode and enter the server URL.

Once connected, you'll see a list of all tools exposed by the MCP server. No extra configuration needed — the Agent will automatically decide when to invoke these tools during conversations.

Each MCP connection has a status indicator next to it: green means connected. If it turns gray or red, click to see details and retry the connection.

Want a quick test? Start with the GitHub MCP server — it lets the Agent view PRs, create Issues, and check CI status directly in conversation, which is incredibly handy.