SpendOS for agent teams.
Control plane for agent operations: governed wallet spend + AI spend governance across IDE, CLI, and agent workflows. Every spend is policy-checked, privacy-preserving, efficiency-aware, and audit-ready.
Live: Frontend · API · PromptOS Terminal
- Policy enforcement — Deny / hold / approve before spend based on budget, vendor, chain, risk.
- Confidential policy evaluation — FHE on Fhenix so budgets and limits stay encrypted end-to-end.
- Confidential vendor selection — Sealed-bid RFP auctions on Canton / Daml with structural sub-transaction privacy and atomic multi-party reveal — see Canton docs.
- Hardware signing — Ledger DMK for high-value transactions with physical confirmation.
- Audit trail — Every decision persisted with evidence, Filecoin + 0G dual-anchor.
- AI safety monitoring — Multi-dimensional suspicion scoring — see report.
pnpm install
pnpm build
pnpm startMinimum .env:
API_KEY=your-api-key-here
OWS_VAULT_SECRET=your-vault-secret
CHAINGPT_API_KEY=your-chain-gpt-key # Optional: Web3 LLM + auditFor local setup, API reference, and production deployment, see the Developer Guide.
31+ Vitest unit/integration tests plus 21 TestSprite CLI backend tests run against the live API (testsprite test run --all --project 8be1ec9e-a2c5-484a-8a2e-422b87832028). Privacy invariants on Canton sealed-bid are asserted by direct ledger queries per party role. CI runs the TestSprite suite on every PR and push via .github/workflows/testsprite.yml. The LOOP documents the write-verify-fix iteration history.
| Doc | What's in it |
|---|---|
| Architecture | System design, data flows, Fhenix / Canton / ChainGPT / Ledger integrations |
| Canton | Daml sealed-bid model, hydration, sandbox + DevNet runbooks |
| Developer Guide | Local setup, API reference, testing, production readiness |
| Deployment | Hetzner / PM2 / nginx, env vars, health checks |
| AI Safety | Suspicion-scoring design, frontier evaluation |
| Hackathon Submission (Canton) | Track 1: Canton private RFP — enterprise framing, DevNet proof pack |
| Hackathon Submission (Arbitrum) | Arbitrum London Founder House — FHE spend governance on Arbitrum Sepolia, agent governance, deployment proof |
| Product & GTM Canvas | Product Canvas + GTM Canvas (Open House London templates) — wedge, why-now, why-onchain, distribution loops |
MIT