Router API reference

Sapiom Router accepts LLM requests from your application using a Sapiom API key. Choose a model, then use the request format your application already understands.

Base URL: https://router.sapiom.ai

Open Router in the dashboard to inspect Router activity. For LLM calls inside a deployed Sapiom agent, see Choose a call surface.

Authentication

Set SAPIOM_API_KEY to your Sapiom API key in the environment where you run the examples. Send it in either Authorization: Bearer or x-api-key. Router authenticates your Sapiom account; you do not supply a provider API key.

Request formats

Format	Endpoint
OpenAI Chat Completions	POST /v1/chat/completions
Anthropic Messages	POST /v1/messages
OpenAI Responses	POST /v1/responses

Each endpoint accepts its named request format and returns that format’s response. Router translates between formats when the selected provider uses a different one.

For Responses requests translated to another format, previous_response_id is not supported. Include the conversation in input instead. Responses requests do not fall back to a different model class.

Models

The models below were verified on September 15, 2026. Availability can change as Router configuration changes.

Use the request ID in the JSON model field with any of the three formats.

Model	Request ID
MiniMax M3	m3
GLM 5.3 Priority	glm-priority
GPT-5.6 Luna	gpt-luna
Claude Opus 4.7	claude-opus
Claude Sonnet 4.6	claude-sonnet

Lanes

A lane selects the latency and capacity policy for your request. Set x-sapiom-lane in any request format. The default is run_now.

Lane	Use
run_now	Interactive work; prioritizes latency.
standard	Less urgent work; balances latency and serving cost.
flex	Delay-tolerant work; accepts slower performance under load.

For example, "model": "m3" with x-sapiom-lane: run_now requests M3 with the interactive policy. Lanes influence routing, capacity admission, and fallback budgets.

flex uses the same request and streaming APIs. It does not submit an asynchronous job. A request can return HTTP 429 if capacity admission fails; a rejected request is not queued for later execution.

Example requests

OpenAI Chat Completions

curl -sS https://router.sapiom.ai/v1/chat/completions \
  -H "Authorization: Bearer $SAPIOM_API_KEY" \
  -H 'Content-Type: application/json' \
  -H 'x-sapiom-lane: run_now' \
  -d '{
    "model": "m3",
    "max_tokens": 256,
    "messages": [{"role": "user", "content": "Say hello."}]
  }'

Anthropic Messages

curl -sS https://router.sapiom.ai/v1/messages \
  -H "x-api-key: $SAPIOM_API_KEY" \
  -H 'anthropic-version: 2023-06-01' \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "claude-sonnet",
    "max_tokens": 256,
    "messages": [{"role": "user", "content": "Say hello."}]
  }'

OpenAI Responses

curl -sS https://router.sapiom.ai/v1/responses \
  -H "Authorization: Bearer $SAPIOM_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "gpt-luna",
    "max_output_tokens": 256,
    "input": "Say hello."
  }'

For streaming, add "stream": true to the JSON body and -N to curl. Router returns a server-sent event stream in the selected request format.