Hi everyone! Thanks for the great work!
I’m running an MCP server using the Docker image ghcr.io/apollographql/apollo-mcp-server:latest. The container spins up successfully. I have an operations folder with 8 operations, and a fairly large GraphQL schema (around 9,000 lines).
When I run the server, the logs show the estimated token size for each tool created. The largest one in my case is 13,665 tokens, with a total across all tools of 31,657 tokens. I’ve also enabled introspection queries.
I’m connecting Claude Desktop (using Claude Sonnet 4.0 with a 200k context window) to this MCP server, but I’m unable to make any queries because the context appears to be too large. Has anyone else run into a similar issue?
I was under the impression that only the tool name and description were exposed to the MCP client/host, so I’m surprised the token count is this high. Any ideas or suggestions would be greatly appreciated—thanks!