The Model Context Protocol (MCP) is an open standard that lets AI assistants and automated agents call our platform data tools natively. This guide outlines how to plug our HTTP streamable server (app.graphiq.ai/mcp) directly into active LLM reasoning loops. Grounding your models inside a pre-resolved graph of 300M+ organizations prunes hallucination loops by up to 87%.
Claude Desktop & Claude Code. Configure a standardized server block in your client config file (.mcp.json), passing keys through an explicit X-API-Key server header wrapper.
For Anthropic Messages pipelines, register our HTTP endpoint inside the beta tools array with the required protocol headers. For OpenAI Responses, configure our secure root URL as an authorized remote tool extension.
LangChain & LangGraph adapters. Expose our discoverable tool logic blocks across custom multi-agent frameworks by deploying standard Python multi-server MCP adapter setups.
How are authentication rights split? Access routes through one of two mechanisms depending on the calling environment.
CLI environments, background scripts, server-to-server networks, and remote LLM provider connectors verify access tokens using an API key header.
Interactive web clients — including claude.ai web views and Claude Desktop app setups — route connections securely via standard browser OAuth 2.0 sign-in loops.
Talk to the grounding assistant to initialize your MCP server connection, run a test query against 300M+ resolved organizations, and verify your agent toolchain configuration end-to-end.
Talk to the grounding assistant →