SpendOS for agent teams.
Cognivern is a control plane for agent operations: governed wallet spend plus AI spend governance across IDE, CLI, and agent workflows.
Move fast without blank checks: every spend can be policy-checked, privacy-preserving, efficiency-aware, and audit-ready.
Live: Frontend · API · PromptOS Terminal
pnpm install
pnpm build
pnpm startMinimum .env:
API_KEY=development-api-key
OWS_VAULT_SECRET=development-ows-vault-secret
CHAINGPT_API_KEY=your-chaingpt-api-key # Optional: Web3 LLM + Smart Contract AuditorSee Developer Guide for full setup and API reference.
Cognivern solves a common agentic scaling problem: teams can ship quickly, but spend control, privacy, and accountability often fragment across tools. The platform unifies those rails so operators can prove disciplined execution without slowing builders down.
| Capability | Endpoint |
|---|---|
| Policy evaluation | POST /api/governance/evaluate |
| Governed spend | POST /api/spend |
| Governed encrypted spend | POST /api/spend/encrypted |
| Spend preview (dry-run) | POST /api/spend/preview |
| OWS wallet bootstrap | POST /api/ows/bootstrap |
| API-key issuance | POST /api/ows/api-keys |
| Audit trail | GET /api/audit/logs |
| Audit permits (confidential) | POST /api/audit/permits |
| Run ledger | GET /api/cre/runs |
| Run ingestion | POST /ingest/runs |
| Natural language intent | POST /api/intent |
| Intent metrics | GET /api/intent/metrics |
- Policy: enforce who/what/when rules before spend executes.
- Privacy: evaluate sensitive policy context via confidential policy paths (
Fhenixadapter). - Efficiency: establish one governance layer for AI operations spend (model/runtime usage visibility and optimization workflows) alongside financial spend controls.
- Audit: persist decision evidence (
decisionId, attestation, run context) for continuous accountability.
# Terminal 1
pnpm start
# Terminal 2
pnpm demo:live| Doc | Covers |
|---|---|
| Hackathon Brief | Demo story, submission, talk track |
| Architecture | System design, data flows, endpoints |
| Developer Guide | Local setup, APIs, testing, production readiness |
| Deployment | Production deployment, PM2, rollbacks |
| Fhenix Integration | Privacy-by-Design plan: encrypted budgets, sealed approvals, FHE policy evaluation |
| ChainGPT Integration | Web3-native AI governance: ChainGPT LLM + Smart Contract Auditor integration |
| Layer | Chain | Contracts |
|---|---|---|
| Execution & Policy | X Layer Testnet (chainId 1952) | GovernanceContract 0x755602bB..., AIGovernanceStorage 0x1E0317bF... |
| Audit Storage | Filecoin Calibration | GovernanceContract + AIGovernanceStorage |
| Confidential Policy State (LIVE) | Fhenix (Sepolia / Arbitrum Sepolia / Base Sepolia) | ConfidentialSpendPolicy + full-stack FHE policy evaluation & authorized decryption — see Fhenix Integration |
| Web3 AI Governance | ChainGPT | Web3 LLM for governance intent + Smart Contract Auditor for runtime defense — see ChainGPT Integration |
Cognivern uses multi-provider AI routing for governance analysis:
| Provider | Use Case | Status |
|---|---|---|
| ChainGPT | Web3-specific governance queries, smart contract analysis | ✅ Primary |
| Fireworks | Cost-efficient general analysis | ✅ Active |
| Kilocode | Free fallback | ✅ Active |
| Workers AI | Cloudflare native | ✅ Active |
| OpenAI | High quality | ✅ Active |
| Gemini | Advanced reasoning | ✅ Active |
| Anthropic | Alternative model | ✅ Active |
The spend control plane is live: policy evaluation, signed approvals, held actions, denials, and persisted run evidence. Contracts deployed on X Layer testnet and Filecoin Calibration. Fhenix integration live for privacy-native operations.
- ChainGPT Integration — Added Web3-native AI governance: ChainGPT Web3 LLM as primary provider with governance context injection, Smart Contract Auditor as runtime pre-spend defense (contract_audit policy rule)
- AI Intent System — Natural language command processing via multi-provider AI routing (7 providers: ChainGPT, Fireworks, Kilocode, Workers AI, OpenAI, Gemini, Anthropic) with circuit breaker protection and graceful fallback
- Compact UI — Reduced visual density across dashboard, cards, and navigation with smooth transitions
- Responsive Improvements — Better mobile/tablet layouts with minimal agent display variants
See Developer Guide for full checklist.
MIT