GraphQL, Explained

What GraphQL actually is, the REST pain it was built to fix, how its typed schema and single endpoint work, and the real trade-offs that tell you when to reach for it and when not to.

1. The Problem REST Leaves REST endpoints over-fetch (they hand you more fields than you asked for) and under-fetch (you need several round trips to build one screen); GraphQL's pitch is to ask for exactly the fields you want and get exactly those, in one request.
2. How GraphQL Works The pieces that make GraphQL run: a typed schema as the contract, one endpoint for everything, queries for reads and mutations for writes, and the rule that the response mirrors the request - shown with an annotated query and its matching JSON.
3. The Real Trade-offs GraphQL isn't free: caching is harder than REST because there's no URL-per-resource, the N+1 problem hits resolvers on the server, flexible queries open the door to complexity and abuse, and you take on extra tooling. When REST or gRPC is the better choice.