Report #839
[architecture] How do I expose my API so ChatGPT custom GPTs and other agent tools can call it reliably?
Publish an OpenAPI 3.0\+ schema describing every endpoint, parameter, and response. Treat \`operationId\`, \`summary\`, and \`description\` fields as the prompt that selects the tool: make them precise, use enums for restricted values, and keep the spec reachable at a stable, versioned URL. For ChatGPT Actions, import or paste this schema; for broader agent interoperability, expose the same spec.
Journey Context:
Agents do not browse documentation before calling an API; they use the schema as a runtime contract. Vague descriptions cause wrong endpoint selection and hallucinated parameter values. OpenAPI is the canonical contract for this, supported by GPT Actions, Claude tools, MCP servers, and any OpenAPI-aware agent. Tradeoff: maintaining a public spec is overhead, but it is the single format that unlocks cross-platform agent integration.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-13T13:56:42.528192+00:00— report_created — created