Documentation
Marturia gives every AI agent decision a cryptographic receipt: signed by your tenant's per-key Ed25519, hash-chained to the previous receipt, and verifiable offline by anyone with the public key. These docs walk through the API, the receipt format, and the open-source verifier.
Quickstart →
Sign up, mint an API key, and emit your first signed receipt in 15 minutes.
API reference →
Endpoint catalogue: OTLP ingest, receipt mint, Article 14 oversight, audit export.
Offline verifier →
pip install marturia-verify — auditors don't need our infrastructure.
Receipt format →
The exact bytes we sign, the hash chain construction, and the JSON canonicalisation rules.
What Marturia is for
EU AI Act Article 12 requires automatic logging of high-risk AI decisions. Article 14 requires verifiable human oversight. ISO/IEC 42001 requires an audit trail that survives system replacement. Marturia generates the cryptographic artefacts that satisfy all three — without making your auditor trust our infrastructure.
What it is not
- Not a model-monitoring tool (latency, drift, hallucinations) — your existing observability stack covers those. Marturia layers on top.
- Not a content-moderation product. Marturia signs whatever decision your agent made; the decision quality is your model's job.
- Not a SOC 2 report. Marturia is evidence; SOC 2 is process. They complement each other.
Help + feedback
Closed beta — fastest support is direct email to [email protected]. Bug reports very welcome. Anything cryptographic that looks wrong gets attention same day.