GraphQL: The Internal Agentic API
summary
Claims that agents can operate every employee workflow by writing GraphQL queries and running them against internal schemas, but many internal endpoints are REST-like so GraphQL power is limited. A five-step pipeline extracts page queries, exposes a subschema, feeds Cloud Code agents the intent, and in four months those agents generate CLI equivalents for most employee actions. Resulting system finds root causes, chains logs/commits, shows emergent end‑to‑end behaviors, but faces N+1 overhead so the team banned the get command and uses GraphQL for deep nested operations.
AI-written summary · about
chapters (5)
- 0:00Introduction and GraphQL history
- 2:31GraphQL for agentic APIs
- 4:40Cloud Code and CLI discovery
- 7:27Emergent behaviors in agents
- 9:53Future of GraphQL and Q&A
related talks
40mGraphQL: A Case Study
19mGraphQL: Accelerating Digital Transformation Through Modern APIs.
37mGraphQL — A New Look at the API
26mJust Ask: Building a Natural Language GraphQL Agent from the Graph You Already…
35mGraphQL in the Enterprise
58mGraphQL -- Flexible APIs for Adaptive Clients
26mOrchestrating APIs for LLMs - Agentic Workflows for GraphQL
29mClosing the Loop: How GraphQL Gives Coding Agents Eyes on What Actually Matters