This document describes the agentic loop as defined by the Open Responses specification and implemented by OGX (the server). Where OGX diverges from or extends the spec, the text is marked with [OGX].
Spec reference: Open Responses specification —
specification.mdxand OpenAPI schemas inschema/.
The Responses API (POST /v1/responses) is an OpenAI-compatible endpoint that supports server-side agentic orchestration. Unlike the Chat Completions API (single inference call), the Responses API runs an iterative loop: the server samples from the model, inspects the output for tool calls, executes internally-hosted tools, feeds results back, and repeats until the model produces a final answer or a limit is reached.