◆AXIOM/inference api

One endpoint. Every model.

OpenAI-compatible chat plus a typed inference protocol — chat, decide, classify, score, predict, rerank, guard. Metered and attributable; the cheapest capable rail is chosen automatically. No hidden routing.

live https://api.axiom.farm

Authentication

Send your key as a bearer token. Use an API key (axp_…, created in the console) for server-to-server calls, or a session token (axs_…) for user-scoped calls.

Authorization: Bearer axp_your_key_here

Keys are shown once at creation. Usage is metered per key and attributable to the key and project that made the request.

Quickstart

curl https://api.axiom.farm/v1/chat/completions \
  -H "Authorization: Bearer $AXIOM_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "axiom-intelligence",
    "messages": [{"role": "user", "content": "Explain the ratchet in one sentence."}]
  }'

Endpoints

POST/v1/chat/completionsOpenAI-compatible chat completion (streaming supported).

POST/v1/inferAXIOM Inference Protocol — a typed task with a typed result.

GET/v1/infer/modelsRegistered inference-protocol models.

GET/v1/modelsModels with customer pricing.

GET/v1/board/{board}Blackboard-as-a-service — shared agent-swarm state.

GET/healthLiveness (public).

Machine-readable

The full OpenAPI 3.1 description is at /openapi.json, suitable for client generators.

Errors

Status	Meaning	What to do
`401`	Missing or invalid key	Check the `Authorization` header.
`402`	Insufficient credits / quota	Top up in the console; the plan quota resets each cycle.
`404`	Unknown model or route	List models at `/v1/models`.
`429`	Rate limited	Back off and retry; honor `Retry-After`.
`502`	Upstream model unavailable	Retry — the router fails over to another capable rail automatically.