What is MCP in OpenClaw?

MCP (Model Context Protocol) is the standard protocol that connects AI models to external tools and data sources. It's the backbone of OpenClaw's skill system and what makes the agent truly powerful.

Think of MCP as a universal adapter. Instead of hard-coding integrations for every service, MCP provides a standardized way for skills to describe their capabilities to the AI model. When a skill is installed, it registers its tools (functions the model can call) and resources (data sources the model can read) with the Gateway.

For example, the google-calendar skill registers tools like 'create_event', 'list_events', and 'delete_event'. When you ask your agent to schedule a meeting, the model sees these available tools and calls the appropriate one with the right parameters.

MCP servers can be written in any language and run as separate processes. This isolation means a buggy skill can't crash your entire agent. OpenClaw manages the lifecycle of MCP servers automatically — starting them when needed and restarting on failures.

The protocol supports three main capabilities: Tools (functions the model can execute), Resources (data the model can read), and Prompts (templates for common interactions). This makes skills portable across different AI platforms that support MCP, not just OpenClaw.

You can find thousands of MCP-compatible skills on ClawHub, or write your own following the MCP specification.

# Install an MCP skill
clawhub install tavily-search

# List installed MCP servers
openclaw skill list