API

Responses API

The OpenAI-compatible /v1/responses endpoint (stateless subset).

POST /v1/responses implements the OpenAI Responses API surface that agents and tools like Codex use. TokenRouter’s implementation is stateless: each call carries its full input.

Supported subset

input as a string or an array of items (message, function_call, function_call_output).
instructions, tools, tool_choice, temperature, max_output_tokens, stream.
Streaming via SSE with typed events (response.output_text.delta, …).

Not supported (stateless): previous_response_id, store, and server-side conversation state. Send the full history in input on every call. Built-in OpenAI tools (web search, file search, code interpreter) are not proxied.

Request

bash

curl https://api.tokenrouter.io/v1/responses \
  -H "Authorization: Bearer tr_your_key_here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-5-mini",
    "instructions": "You are a terse assistant.",
    "input": "List three ways to cut LLM costs."
  }'

Response

json

{
  "id": "resp_7d2f90aa",
  "object": "response",
  "created_at": 1754400000,
  "model": "openai/gpt-5-mini",
  "status": "completed",
  "output": [
    {
      "type": "message",
      "role": "assistant",
      "content": [
        {"type": "output_text", "text": "1. Route to cheaper models... 2. Cap budgets... 3. Cache aggressively."}
      ]
    }
  ],
  "usage": {"input_tokens": 21, "output_tokens": 34, "total_tokens": 55}
}

Function calling

Function round-trips use function_call and function_call_output items in input:

json

{
  "model": "openai/gpt-5-mini",
  "tools": [
    {
      "type": "function",
      "name": "get_weather",
      "parameters": {
        "type": "object",
        "properties": {"city": {"type": "string"}},
        "required": ["city"]
      }
    }
  ],
  "input": [
    {"role": "user", "content": "Weather in Berlin?"},
    {
      "type": "function_call",
      "call_id": "call_abc123",
      "name": "get_weather",
      "arguments": "{\"city\": \"Berlin\"}"
    },
    {
      "type": "function_call_output",
      "call_id": "call_abc123",
      "output": "{\"temp_c\": 21, \"conditions\": \"sunny\"}"
    }
  ]
}

Works with any routed model, including auto — TokenRouter translates to the provider’s native format under the hood.