A colleague and I built this for our own use, but wanted to share it in case folks find it useful.
Why this? GraphQL schemas can easily exceed 75,000 tokens, which makes stuffing the entire schema into an LLM’s context window impractical. Vector‑based RAG often doesn’t help either—chunking the schema leaves the model with partial information. To solve this we taught an agent to traverse the schema graph directly, extracting only the fields and types it needs.
The GitHub repo walks you through wiring the MCP server into Cursor or Claude Desktop and covers the small package‑loading tweak you’ll need.
techno_anthro•7mo ago
Why this? GraphQL schemas can easily exceed 75,000 tokens, which makes stuffing the entire schema into an LLM’s context window impractical. Vector‑based RAG often doesn’t help either—chunking the schema leaves the model with partial information. To solve this we taught an agent to traverse the schema graph directly, extracting only the fields and types it needs.
The GitHub repo walks you through wiring the MCP server into Cursor or Claude Desktop and covers the small package‑loading tweak you’ll need.