OpenAI-compatible · v1

Inference,
with receipts.

Use the client you already have. Turnstile routes each request to the right provider and returns a signed, auditable Compute Pass.

POSThttps://api.turnstileai.net/v1/chat/completions

Make your first request

curl https://api.turnstileai.net/v1/chat/completions \
  -H "Authorization: Bearer $TURNSTILE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openrouter/llama-3.1-70b",
    "messages": [{"role":"user","content":"Explain finality."}],
    "pass": true,
    "anchor": "solana"
  }'

200 OK · 1.84s · pass_8F29KQ issued · signature verified

Request path

One endpoint. Four guarantees.

Route

Your model prefix selects a provider; policy rules can handle fallback.

Measure

Tokens, wall time, provider, and exact cost are captured at the gate.

Seal

Prompt and response hashes are signed into a portable receipt.

Anchor

Optional Solana checkpoints make the proof independently auditable.

Example response

The Compute Pass

pass_8F29KQ● VERIFIED

Modelllama-3.1-70b

Provideropenrouter

Tokens1,204

Latency1,840 ms

Cost$0.0031

Anchor5HkZ9P8m…7nT4

Signature0x8bc902a7…d630

Quickstart →Send a verified completion in five minutes.API reference →Endpoints, schemas, errors, and limits.OpenAI compatibility →Change one base URL in your existing client.