Overview
What is MCP GraphQL Forge?
MCP GraphQL Forge is an MCP server that automatically generates MCP tools from any GraphQL API schema via introspection, enabling AI assistants like Claude or Cursor to interact with GraphQL APIs. It handles parameter validation, field selection caching, and supports both stdio and HTTP transports.
How to use MCP GraphQL Forge?
Install via npx -y @toolprint/mcp-graphql-forge or with Smithery. Set the GRAPHQL_ENDPOINT environment variable (and optionally GRAPHQL_AUTH_HEADER for authentication). Run in stdio mode (default) for direct AI integration, or use --transport http --port 3001 for development/testing.
Key features of MCP GraphQL Forge
- Auto-generates MCP tools from GraphQL schema introspection
- Validates parameters to prevent GraphQL errors
- Supports stdio and HTTP transports
- Caches schema and field selections for performance
- Flexible authentication via environment variables
- Handles circular references in field selections
Use cases of MCP GraphQL Forge
- Connect AI assistants to a Strapi CMS GraphQL API
- Enable Claude or Cursor to query GitHub’s GraphQL API
- Integrate AI tools with e‑commerce GraphQL APIs like Shopify
- Allow natural language queries against any GraphQL endpoint
FAQ from MCP GraphQL Forge
What does it do that a raw GraphQL client does not?
It automatically generates MCP tools from schema introspection, so AI assistants can interact with any GraphQL API without manually writing queries or handling field selection.
What are the runtime dependencies?
Requires Node.js with the TypeScript runtime. Docker runtime support is currently a work in progress.
How does authentication work?
Set the GRAPHQL_AUTH_HEADER environment variable for a Bearer token, or use GRAPHQL_HEADER_* variables for custom headers.
What transports are supported?
The server supports both stdio (for direct AI tool integration) and HTTP (for development/testing with the MCP Inspector).
Does it cache the schema?
Yes. It caches schema and field selections for consistent performance, and you can pre-generate the schema or force regeneration with CLI options.
