The CB Insights MCP Server provides an interface for developers to interact with the CB Insights ChatCBI large language model through AI Agents. It exposes a ChatCBI tool that allows an agent to send a message to ChatCBI and receive the response. Input parameters include: message (the content to send) and an optional chatID to continue an existing conversation. The tool returns a JSON object containing chatID (identifies the conversation), message (ChatCBI's reply), relatedContent (references), sources (sources used to generate the response), suggestions (follow-up questions), and title (the chat title). The server is managed using uv. Configuration is driven by a .env file with CBI_CLIENT_ID and CBI_CLIENT_SECRET for OAuth, along with CBI_MCP_TIMEOUT (default 120) and CBI_MCP_PORT (default 8000). For usage, the README describes integrating with Claude Desktop via mcp install server.py to populate claude_desktop_config.json with a cbi-mcp-server entry pointing to uv and server.py. Debugging can be done with the inspector using mcp dev server.py. The README links to external API docs for additional information on Using ChatCBI.