Designing APIs Both Humans and AI Love
summary
Claim: AI agents now make more API calls than humans and that change requires concrete design changes not vague examples. Use OpenAPI operationId as operationId as tool name, tighten schemas with enums/regex and optional rich fields + strict validation, and expose assessEggEdability workflow endpoint that runs the chain server-side, is deterministic, idempotent, and supports compensating transactions. Logs and metrics show which multi‑step flows to collapse into single workflow endpoints; agents then call one reliable tool instead of brittle multi‑endpoint plans.
AI-written summary · about
chapters (9)
- 0:00AI agents overtake humans
- 1:45Human vs AI expectations
- 5:16AI agents and APIs today
- 8:10Pickable and fillable endpoints
- 10:43AI writes and human reads
- 14:17Simplifying API reads
- 16:22Why AI needs workflows
- 19:56Reliable workflows for AI
- 20:59Key takeaways
related talks
43mDesigning REST APIs for the age of AI agents
24mThe API Economy's Next Chapter: AI Agents as First-Class Clients
35mUnleash Your Inner API Architect: Build Agents That Deliver
40mAPI Lifecycle: Smarter Design and Governance With Open Tooling
26mIf your APIs aren’t agent‑ready, your AI agent is just chat: [...]
27mHow AI is Transforming API Design: Case Studies & Insights
45mFishbowl Discussion on APIs and AI
24mAPIs Are the Nervous System of Agentic AI - Designing Agent-Ready APIs for Aut…