Just Ask: Building a Natural Language GraphQL Agent from the Graph You Already Have (the Easy Way)

GraphQL is powerful, but querying a GraphQL API isn't easy for everyone. Non-developers struggle with schema structure and query syntax. And even experienced developers spend time navigating deeply nested types just to fetch the right data. The promise of “just ask a question and get an answer” feels obvious, yet making it happen has been anything but simple.

Most teams land on RAG pipelines: chunking schema documentation into vectors, embedding it, and retrieving it at query time. It works, but adds unnecessary complexity.

Your GraphQL schema is already a graph. Instead of flattening that structure into vectors and hoping retrieval picks the right pieces, we can store the schema directly in a graph database and let the LLM navigate it.

I'll demonstrate how to model your schema as a graph, build an agent that traverses it to construct and execute queries and return human-readable answers. For common patterns, we'll see how persisted GraphQL queries become reusable tools, making your query library into an agent toolkit.

I'll also introduce Benchmark Broccoli, an open-source project I built that evaluates your mcp tool call performance, running locally.

If you're a GraphQL developer, a platform engineer, or a team exploring LLM-powered APIs, this talk gives a practical, simpler architecture you can start building with today.