SentinelPay is a secure, on-chain infrastructure layer for autonomous AI agents on the Celo network. It separates the "brain" (Agent Logic) from the "wallet" (Financial Authority), ensuring AI agents cannot go rogue and overspend or pay unauthorized addresses

SentinelPay is a security-first infrastructure layer for autonomous AI agents on the Celo blockchain.Its core mission: Provide a trustless, deterministic way for AI agents to execute payments (e.g. in USDC) without directly holding private keys or being able to overspend/misbehave.
Key points in one line:

1. Separates AI logic (off-chain) from financial authority (on-chain via SentinelVault smart contract).
2. Enforces spending policies, limits, and whitelists immutably on-chain.
3. Builds on ERC-8004 "Trustless Agents" standard for agent identity + reputation.
4. Solves risks like policy drift, key leaks, and lack of verifiable agent behavior.

AI agents are rapidly becoming autonomous, but they lack standard, trustless guardrails. Giving an agent a private key is a "blind trust" model, leading to risks like:

1. Policy Drift: AI logic errors causing overspending.
2. Security Breach: Private keys being leaked from agent runtimes.
3. Opacity: No verifiable on-chain reputation for "good acting" agents.

SentinelPay adds an immutable on-chain policy gate via the SentinelVault contract. It maps to the ERC-8004 "Trustless Agents" standard to give agents a verifiable identity and reputation.

Our Innovation:

1. Deterministic Enforcement: Limits are enforced by Celo smart contracts, not just backend logic.
2. ERC-8004 Native: Built-in support for agent identity and reputation tracking.
3. Defense in Depth: HMAC signing, idempotency protection, and on-chain whitelisting.