The Verified Fact Stamp about your agent — compliance and regulation conformity, anchored on chain.
Paste an ERC-8004 agent URL or a GitHub repo. We resolve the source code, audit it against the EU AI Act and NIST AI RMF, and publish a verifiable score on chain. AI is global; regulation is local; AiAuditor makes compliance portable and the audit trail verifiable.
https://8004scan.io/agents/ethereum/9382 → per-clause audit → on-chain attestation
Status: V0 (Day 0) — landing page + agent resolution shipping today. Full pipeline in active build.
The EU AI Act becomes fully applicable on 2 August 2026. Penalties reach €35M or 7% of global turnover. ERC-8004 is already past 45,000 registered AI agents as of 2026. None of them have a standard verifiable compliance record. AiAuditor is that record.
Multi-stage pipeline that adapts established security-audit patterns to regulatory clauses:
INTAKE → FETCH → RECON → SCOPE → MAP → CHECK → VERIFY → GRADE → REPORT → ANCHOR
Each in-scope clause is checked by a deterministic rule and (where needed) an LLM judge. Every finding cites file and line. The final score is anchored on Sepolia as a packed AuditScored event plus a canonical ERC-8004 validationResponse.
- Next.js 16 (App Router) + viem.
- Landing page with intake form.
/api/resolve— POST an 8004scan.io URL, get the resolved agent metadata + repository URL./a/{chain}/{tokenId}— agent detail page with waitlist for the full audit.- Vercel-deployable as-is.
git clone https://github.com/drhus/ai-auditor.git
cd ai-auditor
npm install
cp .env.example .env.local
# fill in registry addresses + RPC URLs
npm run devThen visit http://localhost:3000.
vercel link
vercel env pull
vercel --prodThe end-state is continuous attestation — every commit triggers a new audit, every score is live on chain, every agent's compliance trail evolves with the code. Full detail in docs/vision.md.
- V0 (today): landing + agent resolution.
- V0.5 (2 weeks): repo cloning + first 5 EU AI Act clause checkers (Art 12, 14, 15, 50; high-risk classification).
- V1 (4–6 weeks): Sepolia contract deployed, full pipeline live, public directory of audits, 20+ clauses across EU AI Act + NIST AI RMF.
- V2 (+8–12 weeks): GitHub App + CI/CD — re-audit on every commit, score evolves with the code.
- V3 (+3–4 months): MCP server — Claude Code / Codex / autonomous agents audit themselves before deploy.
- V4 (+4–6 months): Multi-jurisdiction — one audit, per-market scores for EU + US + UK + Japan + Singapore + more.
- V5 (+6–12 months): Decentralised audit market on ERC-8004's Validation Registry.
Full project documentation lives in docs/:
- Overview · PRD · Team brief
- Pipeline design · On-chain anchoring · User journey
- Regulations matrix · V0 MVP spec
MIT.