> ## Documentation Index
> Fetch the complete documentation index at: https://docs.trynito.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# The Attestation Pipeline
> The end-to-end flow of a Confidential attestation, from the nonce you supply to everything that comes back.
Fetching an attestation runs a fixed sequence. The [seven-gate baseline](/privacy/tee-attestation/seven-gate-baseline), [the report-data binding](/privacy/tee-attestation/freshness-binding), and [per-provider behavior](/privacy/tee-attestation/per-provider-attestation) all describe steps inside this flow.
## What You Provide
| Input | Required | Purpose |
| -------------------------- | -------- | ------------------------------------------------------------------------------------------ |
| The Confidential model | Yes | The model to attest. It must resolve to an attestation-capable sealed-compute environment. |
| A fresh value (nonce) | No | Binds the report to this request. Omitting it permits an empty binding. |
| Expected signing address | No | The signing address you expect the request binding to commit to. |
| Expected signing algorithm | No | If given, it must match what the evidence declares. |
The nonce is optional, but supplying a new random value every request is what gives you replay protection. Anti-replay is caller-driven and no past values are stored, so an omitted nonce is permitted and produces an empty-binding report. See [the report-data binding](/privacy/tee-attestation/freshness-binding).
## What Nito Does
1. **Resolve the model** to a concrete, attestation-capable sealed-compute environment. A model that does not support attestation has nothing to attest.
2. **Fetch the hardware evidence** from that provider. How each one frames the request differs: see [per-provider attestation](/privacy/tee-attestation/per-provider-attestation).
3. **Run the baseline verifier** over the evidence. Those checks set whether the result is reported as baseline-verified.
## What Comes Back
A structured summary alongside the raw evidence:
* The model you asked to attest, and the sealed-compute environment it resolved to.
* The privacy level recorded for that model, and whether attestation is available for it.
* The subject the provider was asked about, where the provider frames its request that way.
* **Whether the baseline passed.**
* The nonce observed in the evidence, to compare against the one you sent.
* The proof format and signing algorithm the evidence declares.
* The signing address committed to in the request binding, and the model public key.
* **The raw hardware evidence in full**, for your own verification.
The summary reports the model public key once. Some tooling surfaces it under a second name such as "signing key," but it is one value, not two independent keys.
Attestation results are request-local and are not cached. Each request produces a fresh report bound to the nonce you sent, so an old report cannot be replayed as a new one.
## The Scope Limit
A passing baseline means the evidence passed Nito's baseline checks. It does **not** mean Nito validated the underlying hardware quotes, the certificate chains, the revocation lists, or the Trusted Computing Base policy. Those are deliberately out of scope and are yours to run on the raw evidence, which is returned in full for exactly that reason. See [verify it yourself](/privacy/tee-attestation/verify-yourself-recipe).
## Related Resources
The exact checks Nito runs over the evidence.
How your nonce is bound in and compared.
How evidence differs across backends.