Get attestation

GET /api/v1/attestation

The gateway fetches the evidence from the model’s sealed backend, binds a challenge, checks the envelope, and returns it raw. Nito does not verify the hardware signatures: verified is always false. The same fetch runs, with a gateway-generated challenge, before every Confidential chat call. This endpoint is how you hold evidence bound to a challenge you chose, and where an Encrypted client gets the model public key. See How Attestation Works and Verify yourself.

Query Parameters

Key	Required	Notes
`model`	yes	A logical model ID with the `:confidential` suffix, such as `z-ai/glm-5.2:confidential`. The model must appear in `GET /v1/models` with a `:confidential` ID.
`trust_tier`	no	`confidential` (the default) or `encrypted`. With `encrypted`, the model must be an `:encrypted` entry and the evidence must carry a model public key of the right shape for the declared algorithm, and the response reports `privacy_tier: "encrypted"`.
`nonce`	no	Your challenge: exactly 32 random bytes as 64 hexadecimal characters. When omitted the gateway generates one. It is request-local and never stored.
`signing_address`	no	Provider-specific signing address to forward to the adapter. Request-local.
`signing_algo`	no	Provider-specific signing algorithm to forward. Request-local.

Unknown parameters return 400. Unknown, non-attested, or below-Confidential models are rejected before any provider call.

Response

{
  "object": "attestation",
  "model": "z-ai/glm-5.2:confidential",
  "provider_model_id": "...",
  "privacy_tier": "confidential",
  "attestation_available": true,
  "attestation_subject": "...",
  "verified": false,
  "verifier_provenance": "",
  "nonce": "0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef",
  "proof_format": "...",
  "signing_algo": "ed25519",
  "signing_address": "...",
  "signing_key": "...",
  "model_public_key": "...",
  "evidence": { "quote": "provider-specific request-local evidence" }
}

Field	Notes
`object`	Always `attestation`.
`model`	The model you requested.
`provider_model_id`	The backend’s native model binding used for attestation.
`privacy_tier`	`confidential`.
`attestation_available`	Whether evidence could be gathered.
`attestation_subject`	The attested subject the backend reported.
`verified`	Always `false`. The gateway does not verify evidence.
`verifier_provenance`	Always empty. Kept for wire compatibility.
`nonce`	The challenge, always populated: yours if you sent one, otherwise the generated value. Expect it inside the evidence’s report data.
`proof_format`	The backend’s proof format identifier: the evidence layout and how the challenge and keys are bound.
`signing_algo`, `signing_address`, `signing_key`, `model_public_key`	As observed from the backend.
`evidence`	The raw evidence, provider-specific, for independent verification: the Intel TDX quote and, where the backend runs a GPU, the NVIDIA GPU evidence.

Served with Cache-Control: no-store, never persisted. Fetch fresh evidence with a fresh challenge each time.

What the Gateway Enforces

It rejects unsupported parameters, unknown or non-attested models, and empty evidence, and returns 503 unavailable when the backend echoes a challenge that contradicts the one it was given.

Example

NONCE=$(openssl rand -hex 32)
curl "https://app.trynito.ai/api/v1/attestation?model=z-ai/glm-5.2:confidential&nonce=$NONCE" \
  -H "Authorization: Bearer z_sk_your_key_here"

Basics

Features

Server-Side Tools

Multimodal

Best Practices

SDKs

API Reference

Get attestation

Query Parameters

Response

What the Gateway Enforces

Example

​Query Parameters

​Response

​What the Gateway Enforces

​Example

​Related

Query Parameters

Response

What the Gateway Enforces

Example

Related