Project AIR
vindicara.io / design partners
View source · MIT $ pip install projectair
Design Partner Program

Become one of three founding design partners.

Project AIR is the system of record for human accountability over AI agents. It monitors every action, enforces a deterministic floor an uncovered agent cannot cross, proves the result with signed and anchored evidence, and ties every action back to a named human. Evidence-grade infrastructure is the foundation, not the whole story.

We will integrate Project AIR into your production agents, hands-on and free. In return, we ask for structured feedback and an anonymized case study. You keep direct influence over the roadmap and can verify the technology before committing.

Prove it before you sign anything
  • pip install projectair then air demo: a real signed chain in under 30 seconds, no account, offline.
  • Verify our own production ops chain yourself on public Sigstore Rekor. Zero Vindicara infrastructure in the verification path.
  • Read the admissibility architecture, designed for FRE 902(13) self-authentication and eIDAS mapping with honest disclosures, at /admissibility.
What you get
White-glove SDK integration
We wire Project AIR into your agents across major frameworks: LangChain, OpenAI, Anthropic, LlamaIndex, Gemini, and Google ADK.
Cryptographically signed, replayable forensic chain
For every agent session, with 16 detectors across the OWASP Top 10 for Agentic Applications, OWASP LLM, and AIR-native checks.
Deterministic enforcement floor
An uncovered agent cannot cross it. Enforcement runs as fixed logic, not a model that can be jailbroken.
Compliance-ready exports
EU AI Act Article 72 and Article 12 artifacts, plus HIPAA audit-control evidence (45 CFR 164.312(b)).
Direct line to the founder
Real influence over the roadmap, and a BAA where PHI is involved.
What we ask
A production or near-production deployment
A real agent workload we can instrument.
Two or three short feedback sessions
Over the course of the partnership.
A case-study commitment
Anonymized at launch is fine; named when you are comfortable, plus a quote we can use.
Why now
EU AI Act Article 72
Applies August 2, 2026 for high-risk systems under Annex III (HR, credit, essential services, and more), requiring continuous post-market monitoring you can hand to a regulator.
HIPAA is tightening
It already requires audit controls (45 CFR 164.312(b)); the 2025 Security Rule NPRM (proposed) signals stricter requirements ahead.
Your agents are in production now
And standard logs are mutable, gapped, and cannot prove they were not tampered with.
The outcome

Independently verifiable, tamper-evident evidence for every agent decision, with enforcement that holds and authority that traces to a person. Built for regulators, auditors, customers, and your own risk team.

Free during the design-partner period. BAA available. About a 30-minute kickoff to start.