GLACIS Documentation

Zero-egress AI attestation. Create cryptographic proofs of AI operations without sensitive data leaving your environment.

Get Started View on GitHub

Install

pip install glacis

Quick Example

from glacis import Glacis

# Offline mode - works immediately, no API key needed
glacis = Glacis(mode="offline")

receipt = glacis.attest(
    service_id="my-ai-app",
    operation_type="inference",
    input={"prompt": "..."},   # Hashed locally, never sent
    output={"response": "..."}, # Hashed locally, never sent
)

print(receipt.attestation_id)  # oatt_xxx
print(receipt.witness_status)  # UNVERIFIED

What Gets Sent?

Data	Transmitted?
Your prompts	No - SHA-256 hash only
Model outputs	No - SHA-256 hash only
API keys	No
Timestamps	Yes
Service ID	Yes

Your sensitive data stays local. Only cryptographic commitments are transmitted for witnessing.

Zero-Egress

Payloads are hashed locally using SHA-256. Only the hash is transmitted - your data never leaves your environment.

Offline Mode

Works immediately without an API key. Receipts are self-signed with Ed25519 and marked “UNVERIFIED”.

Provider Integrations

Auto-attesting wrappers for OpenAI and Anthropic. Every API call is automatically attested.

Verifiable

Receipts can be independently verified. Online mode adds Merkle proofs and a transparency log.

SDK Documentation

Installation Install the SDK with pip, poetry, or uv.

Quickstart Create your first attestation in 3 lines of code.

Offline Mode Use the SDK without an API key for development.

OpenAI Integration Auto-attest every OpenAI API call.

Offline vs Online Mode

Feature	Offline	Online
API key required	No	Yes
Signing	Local Ed25519	Glacis witness
Merkle proofs	No	Yes
Verification URL	No	Yes
Witness status	UNVERIFIED	VERIFIED

Start with offline mode for development. Upgrade to online when you need third-party verifiability for audits, papers, or customer due diligence.