cbi

The CB Insights MCP Server provides a pass-through interface to the ChatCBI LLM, exposing a simple API to send messages to ChatCBI and receive structured responses including chat identifiers, content, sources, and suggested follow-ups. It leverages uv to host the Python server and reads configuration and credentials from environment variables, enabling secure OAuth-based access to CB Insights services. You can use the built-in ChatCBI tool to initiate or continue conversations, retrieve related content, and gather references and recommendations for downstream agents or workflows. The server is designed to be drop-in compatible with other MCP clients that speak the common MCP protocol, so you can integrate it into your agent pipelines or testing tools with minimal changes.

To interact with the server, ensure you have the required environment variables configured (client ID/secret for CB Insights, and network port settings). Start the server using your chosen orchestration method, then send requests to the exposed MCP endpoint to invoke ChatCBI. The response payload includes fields such as chatID, message, relatedContent, sources, suggestions, and title, which you can display or log in your UI or agent reasoning loop.

Prerequisites:

• Python 3.8+ installed
• uv (Uvicorn-like runner) installed in your environment
• Access credentials for CB Insights API (CBI_CLIENT_ID and CBI_CLIENT_SECRET)

Step-by-step:

1. Clone the MCP server repository: git clone <repo-url> cd <cloned-repo>

2. Create and configure a .env file or export environment variables: export CBI_CLIENT_ID="<your-client-id>" export CBI_CLIENT_SECRET="<your-client-secret>" export CBI_MCP_TIMEOUT=120 export CBI_MCP_PORT=8000

3. Install Python dependencies (if a requirements file exists, e.g., requirements.txt): python -m pip install -r requirements.txt

4. Start the server with uv: uv --directory /path/to/cloned/cbi-mcp-server run server.py

5. Verify the server is running by hitting the MCP endpoint, for example: curl http://localhost:8000/mcp/health

6. If you are using Claude Desktop or another client, configure its mcpServers entry to point to the running uv instance as shown in the example configuration.

Environment variables:

• CBI_CLIENT_ID and CBI_CLIENT_SECRET are required for authenticating with CB Insights.
• CBI_MCP_TIMEOUT controls the request timeout (default 120 seconds).
• CBI_MCP_PORT defines the listening port (default 8000).

Common issues:

• If the server cannot authenticate, double-check the OAuth credentials and ensure the client is authorized for API access.
• Ensure network egress to CB Insights API is allowed from your runtime environment.
• When updating credentials, restart the uv server to pick up changes.

Debug tips:

• Use the inspector tool via the MCP to test endpoints and inspect responses.
• Use the provided mcp install/dev commands to validate configuration files before production use.