Responses API

OpenAI Responses-compatible endpoint for newer clients, agents, reasoning, tool calls, and multimodal workflows when supported upstream.

POSThttps://onerouter.com/v1/responses

Troubleshooting

Errors carry an HTTP status and a JSON error body. See Errors & Retries for the full list and retry guidance.

Supported models

Works with models that expose the newer OpenAI Responses object — including reasoning models. Browse the Text & Chat Models page or call GET /v1/models.

Overview

The Responses endpoint is compatible with OpenAI's Responses API — the newer object model designed for agents, reasoning, and multi-step output. Unlike Chat Completions' messages array, you send input (a string or structured items) and read the model's output from the output array. Reasoning summaries, tool-call items, and multi-step events appear as typed items in that array.


Use Responses when your client or workflow expects this object shape. Use Chat Completions when you need the broadest SDK compatibility.

Request headers

ParameterTypeRequiredDescription
Authorizationstring · requiredrequiredBearer token, formatted as `Bearer YOUR_API_KEY`.
Content-Typestring · requiredrequiredMust be `application/json`.

Request body

ParameterTypeRequiredDescriptionValues
modelstringrequiredModel ID in `provider/model` form.
inputstring | arrayrequiredUser input as a string, or an array of structured input items.
instructionsstringoptionalHigh-level system-style behavior instructions.
toolsarrayoptionalTool definitions, including web search or function tools when supported.
reasoningobjectoptionalReasoning config for reasoning models, e.g. `{"effort": "medium"}`.
text.formatobjectoptionalStructured text output format (JSON schema) when supported.
max_output_tokensintegeroptionalCap on generated output tokens.
streambooleanoptionalEmit response events as SSE.
truefalse

The output array

Read results from output[]. A simple text reply is a message item with content[].type === "output_text". Reasoning models may add a reasoning item; tool use appears as function_call items. Most SDKs expose a convenience accessor (e.g. response.output_text) for the plain text.

Model-dependent features

OneRouter keeps protocol compatibility and forwards model-specific parameters where supported. Features such as tool calling, structured output, reasoning, prompt caching, web search, image input, audio input, video generation, and provider-specific options depend on the selected model and upstream provider. Check the model detail page before relying on a feature in production.