🤖🛡️ AgentPolicyKit (Celo)

Autonomous agents with handcuffs.

Permission + spending control layer for AI agent payments on Celo

---

⚡ TL;DR

Most agent stacks force a bad choice:

• give your bot full wallet control (unsafe) ❌
• approve every transaction manually (not autonomous) ❌

AgentPolicyKit gives a third option:

✅ programmable policies onchain
✅ least-privilege execution
✅ instant revoke / pause
✅ auditable logs for every action

---

🧠 One-liner

AgentPolicyKit is a security-first middleware (contracts + SDK) that lets AI agents execute onchain payments on Celo under verifiable policies (limits, allowlists, time windows, and instant revocation).

---

🏗️ What’s inside

contracts/
  ├─ PolicyRegistry.sol      # Create / update / revoke agent policies
  └─ ExecutionGuard.sol      # Validate each action before execution

sdk/
  └─ TypeScript SDK          # createPolicy, executeWithPolicy, revokePolicy

demo/
  └─ Autonomous Payroll Agent# cUSD payouts under policy constraints

docs/
  ├─ ONE_LINER_SUBMISSION.md
  ├─ ARCHITECTURE_DIAGRAM.md
  ├─ BUILD_CHECKLIST_48H.md
  └─ PITCH_DECK_5_SLIDES.md

---

🔐 Security model (human language)

• No blind trust in agents
• Fail-closed validation (if policy check fails, tx fails)
• Scoped permissions (amount, recipient, method, time)
• Emergency controls (pause/revoke now)
• Auditability by default (allow/deny events)

---

🔄 Architecture

flowchart LR
    O[Owner / Multisig] -->|createPolicy/revokePolicy| PR[PolicyRegistry.sol]
    A[Agent Runtime] -->|executeWithPolicy| EG[ExecutionGuard.sol]
    PR -->|policy lookup| EG
    EG -->|if valid| T[Target contracts on Celo]
    EG -->|emit logs| L[(Audit Events)]

Loading

---

🧪 Real-world use cases

• 💸 contributor payroll in cUSD
• 🏦 treasury ops with daily limits
• 🛒 commerce automations (whitelisted merchants)
• 📅 recurring payouts under policy windows

---

🏆 Why this fits Celo Infra Track

Criterion	How AgentPolicyKit scores
Technical Innovation	Onchain policy engine for autonomous agent finance
Developer Experience	Minimal SDK + templates for common payment flows
Security & Trust Minimization	Least privilege, revocation, audit logs
Real-World Applicability	Immediate use in payroll/treasury/agent commerce

---

🚀 48h build plan

See: docs/BUILD_CHECKLIST_48H.md

Quick milestones:

• contracts v1
• SDK v1
• payroll demo
• 2-min demo video
• Karma + submission assets

---

📚 Submission docs

• ONE_LINER_SUBMISSION.md
• ARCHITECTURE_DIAGRAM.md
• BUILD_CHECKLIST_48H.md
• PITCH_DECK_5_SLIDES.md

---

✊ Build philosophy

Autonomous systems should increase human agency, not wallet risk.

No hype wrappers. No trust-me security.
Just clear policy rails for bots that move real money.

---

If your agent can pay, your policy layer should say how.