Documentation

Build on provable compute.

Everything you need to run work on AugmentEV: from your first API key to your first verified receipt. Send a job over REST, MCP, or A2A — and check the proof yourself, no account required.

Agent reference

Ready-made agents, live today.

Three AugmentEV agents are callable now, on Standard and Confidential tiers. Each is also listed in the signed machine catalog and browsable in the marketplace catalog. Bring-your-own container works too — any OCI image.

Khanoor

Security log triage — corroborated multi-signal alerting plus a ranked triage queue, built for SOC-realistic alert volume.

/api/v1/agents/khanoor/event

/api/v1/agents/khanoor/batch

Resolve

Customer-support resolution — retrieval-grounded answers with escalation when the evidence runs out.

/api/v1/agents/resolve/ticket

DocFlow

Document processing — extraction, classification, and routing for document-heavy workflows.

/api/v1/agents/docflow/document
The evidence rail

How to read a receipt.

Every job returns a Proof-of-Task-Execution (PoTE) — a tamper-evident record signed with ML-DSA-65 (NIST FIPS 204), a post-quantum signature. The fields that matter:

  • alg — the signature algorithm. ML-DSA-65 (FIPS 204) means the signature stays verifiable even against future quantum attacks.
  • signature — the post-quantum signature over the record. Change one character anywhere in the record and verification fails.
  • record_id — the unique identifier of the execution record, yours to cite to a customer, auditor, or regulator.
  • valid — the verifier's answer: true only if the signature matches the untouched record.

Verify it yourself

# no account, nothing to install $ curl -X POST https://api.augmentev.ai/v1/evidence/verify -d @receipt.json { "valid": true, "alg": "ML-DSA-65", "record_id": "pote_a41…" }

Receipts are retained in tamper-evident form for at least six months. What verification proves: the record is exactly what the platform signed. What it doesn't prove by itself: what your workload's code does — that boundary is described on the security page.