apollo-mcp-server

Guide for using Apollo MCP Server to connect AI agents with GraphQL APIs. Use this skill when: (1) setting up or configuring Apollo MCP Server, (2) defining MCP tools from GraphQL operations, (3) using introspection tools (introspect, search, validate, execute), (4) troubleshooting MCP server connectivity or tool execution issues.

83stars10forksUpdated 6/18/2026

MCP Tools API Development #api-integration #apollo #graphql #mcp #ai-tools #schema-introspection

Security Assessment

Low Risk(75/100)

Detected risks:

Remote Code Execution([SKILL.md] curl -sSL https://mcp.apollo.dev/download/nix/latest | sh)

Security Score75/100

About apollo-mcp-server

Apollo MCP Server is a bridge that connects AI agents to GraphQL APIs through the Model Context Protocol (MCP). It exposes GraphQL operations as callable MCP tools, enabling AI assistants like Claude to interact with GraphQL endpoints without requiring custom integration code. The server handles schema introspection, operation validation, and query execution, making GraphQL APIs accessible to AI agents in a standardized way.

The server provides four built-in introspection tools for exploring GraphQL schemas: introspect for detailed type definitions, search for finding specific types and fields, validate for checking operation validity, and execute for running ad-hoc queries. Custom tools can be defined through three methods: local operation files where each GraphQL operation becomes an MCP tool, operation collections managed through GraphOS Studio for collaborative workflows, and persisted query manifests for production environments with pre-approved operations.

This skill is designed for developers integrating AI agents with GraphQL APIs, teams building AI-powered tools that need to query GraphQL endpoints, and organizations wanting to expose their GraphQL APIs to AI assistants securely. It works with Claude Desktop, Claude Code, and Cursor, supporting both streamable HTTP transport for network-based deployments and stdio transport for direct client launches. The server is particularly useful for enabling AI agents to perform schema exploration, validate queries before execution, and interact with GraphQL APIs through natural language conversations.

FAQ

What transport methods does Apollo MCP Server support?

Apollo MCP Server supports two transport methods: streamable HTTP (recommended) where the server runs independently and clients connect via HTTP at http://127.0.0.1:8000/mcp by default, and stdio transport where the MCP client launches the server directly as a subprocess.

How do I define custom MCP tools from GraphQL operations?

You can define custom tools three ways: (1) Local operation files - place GraphQL operation files in a directory, each file with exactly one named operation becomes a tool; (2) Operation collections - reference a GraphOS Studio collection ID; (3) Persisted queries - use a persisted query manifest JSON file for production environments.

Which AI clients are compatible with Apollo MCP Server?

Apollo MCP Server works with Claude Desktop, Claude Code, and Cursor. Configuration varies by client - Claude Desktop uses claude_desktop_config.json, Claude Code uses the 'claude mcp add' command or .mcp.json, and all clients can connect via npx mcp-remote for streamable HTTP transport.

What are the four built-in introspection tools?

The four built-in tools are: introspect (explore schema types in detail with full type definitions and relationships), search (find specific types or fields in the schema), validate (check if GraphQL operations are valid before execution), and execute (run ad-hoc GraphQL queries and mutations for testing).

How do I configure the GraphQL endpoint location?

The GraphQL endpoint defaults to http://localhost:4000/. If your API runs at a different location, override this by adding an 'endpoint' key in your config.yaml file. The MCP server endpoint for streamable HTTP defaults to address 127.0.0.1 and port 8000.