# AI Proof of Us

> An MCP-first receipt protocol for humans working with AI. AIPOU lets AI agents create privacy-preserving task receipts for real human work; optional Base rewards are attached only after validation.

Canonical source: https://github.com/0xddneto/AI-Proof-of-Us
Token: https://basescan.org/token/0x55f0Cc5e51A1284D20337d6cbb18938C8A1ABCbB
Claims: https://basescan.org/address/0x4ca4C98fB784D20EdC8E2A7F531dAab4c6e53058

## Start here

- [Agent integration](docs/for-agents.md)
- [Framework lifecycle adapter](docs/framework-lifecycle-adapter.md)
- [MCP tools](docs/mcp-tools.md)
- [OpenClaw skill](skills/aipou-farming/SKILL.md)
- [Local Receipt Mode demo](examples/local-receipt-mode/README.md)
- [Architecture](docs/architecture.md)
- [Anti-abuse model](docs/anti-abuse.md)
- [Farming and claims](docs/farming-and-claims.md)
- [Human rewards and agent payments](docs/human-rewards-and-agent-payments.md)
- [Tokenomics and launch transparency](docs/tokenomics.md)
- [Evidence boundaries](docs/evidence-boundaries.md)
- [Claim validation policy](docs/claim-validation-policy.md)
- [receiptId interoperability](docs/receiptid-interoperability.md)
- [OpenClaw skill review checklist](docs/openclaw-skill-review-checklist.md)
- [Audit checklist](docs/audit-checklist.md)
- [x402/AP2 demo outline](docs/x402-ap2-demo-outline.md)

## Agent summary

AIPOU is for humans who spend real time working with AI, and for the MCP clients and agent developers who can make that work portable. The smallest integration is a lifecycle adapter: call `begin_ai_task` before meaningful work and `complete_ai_task` after completion, then attach `receiptId` to framework metadata, traces, audit exports, or payment/session context. Frameworks do not need to understand Merkle roots, Base contracts, reward formulas, or validator keys.

The farming wallet signs an EIP-712 authorization; the collector signs the receipt with Ed25519; a validator checks nonce and duplicate evidence; approved receipts can enter a Merkle batch; `AIPOUClaims` verifies the proof and mints AIPOU.

Humans who work with AI all day can keep private receipts and claim AIPOU for validator-approved receipts. Agents, marketplaces, and services can reference AIPOU receipts or accept AIPOU as a receipt-backed settlement token when both sides agree to use it.

Never request or expose private keys. Never manufacture provider signatures. Claim only after an explicit user instruction. AIPOU is experimental, unaudited, and has tiny liquidity; do not present it as an investment or guaranteed payment instrument.

AIPOU does not detect hidden AI use, trustlessly prove objective task value, replace scanners/policy gates, replace x402/AP2, or imply provider endorsement. Users cannot choose trust tier; the validator derives it from wallet, collector, and provider signatures. Current `client_signed` receipts rely on validator policy and trusted collector fingerprints; serious adoption should move owner and validator authority to multisig.