apollo-mcp-server

Official
Local
258
GitHub Repo

Overview

The apollo-mcp-server is a Model Context Protocol (MCP) server implementation that makes your GraphQL API operations available as MCP tools for AI assistants and agents. It acts as a translation layer between AI clients and your GraphQL infrastructure, exposing query and mutation capabilities through MCP so models can fetch, explore, and orchestrate your graph operations without bespoke tooling.

Transport

streamable-http

Tools

The server converts GraphQL operations into a set of callable MCP tools such as:

  • execute – Run a specific GraphQL query or mutation against your API
  • introspect – Retrieve schema metadata and structure
  • search – Locate types, fields, or operations in the schema
  • validate – Check GraphQL operation syntax before execution

These tools are generated from your defined operations or schema introspection, enabling AI clients to meaningfully interact with the GraphQL API.

Key Capabilities

  • Expose GraphQL operations as MCP tools, allowing AI agents to query and mutate your API programmatically.
  • Support multiple transport protocols including STDIO and Streamable HTTP, giving flexibility for local development or cloud deployments.
  • Enable schema exploration via introspection, so AI assistants can discover data models and operations dynamically.
  • Use persisted queries or schema-driven tool generation, providing governance over which capabilities are exposed.
  • Integrate with MCP clients such as Claude Desktop, ChatGPT, Cursor, and others, so your GraphQL APIs confidently become part of AI workflows.

How It Works

The apollo-mcp-server runs alongside your GraphQL API and exposes an MCP endpoint that AI clients can connect to using supported transports such as STDIO or HTTP/SSE. At startup, the server processes your GraphQL schema and operation definitions — either from local .graphql files or from a GraphOS persisted query manifest — and generates a corresponding set of MCP tools.

When an AI agent invokes one of these tools, the MCP server executes the associated GraphQL operation against your API endpoint, then returns structured results over the MCP protocol. This design lets AI clients fetch data, run mutations, inspect the schema, and integrate GraphQL operations into larger automated reasoning flows without custom integration code.

Deployment options include running locally via Apollo’s Rover CLI for development or deploying in containers (including the Apollo Runtime container) for production at scale, while still respecting your existing authentication, headers, and security policies.