Commit Protocol - Built for MNEE

Optimistic settlement for on-chain work

Read the whitepaper here : https://github.com/ayush18pop/mnee-commit/blob/main/commit-protocol/commit_protocol.pdf

Inspiration

Most work agreements fail for one reason: payment is discretionary.

Live on Sepolia

The protocol is currently deployed and verified on the Sepolia Testnet. The system has been rigorously tested on a forked Anvil chain, ensuring it is ready for mainnet deployment.

Commitment Contract: 0x63c0eA0260846bAaF4b73c3792699bf205B6c3ba

MNEE Token Address: 0xE9cBe461c30021257c52da81674B38158DD23FF5

In DAOs, Discord communities, and open collaboration, payment depends on someone remembering, approving, or choosing to pay. This introduces delay, bias, negotiation, and in many cases non-payment. Even with smart contracts, most systems still replicate the same human bottleneck: a multisig, an admin, or a final approval step.

We asked a different question: What if payment was not a decision at all? What if settlement was the default state of a commitment, and humans only had the ability to intervene at a cost?

This led to Commit Protocol: a system where time executes payment, AI prices interference, and economic incentives enforce honesty. Not a marketplace. Not an app. A financial primitive for work.

What it does

Commit Protocol is a financial automation protocol for irreversible work commitments. It allows Discord communities, DAOs, and teams to create work commitments where:

Funds are escrowed at creation
There is no “Pay” button
Settlement happens automatically after a deadline
Disputes are possible but economically expensive
AI agents do not approve work and never release funds

Instead, AI only increases the cost of lying.

Core Behavior

A Discord server registers once and maintains a prepaid MNEE treasury.
Authorized roles create commitments and funds move into escrow immediately.
Contributors submit verifiable evidence such as code, artifacts, or specs.
A settlement clock starts.
If no one stakes capital to stop it, payment is inevitable.

The system defaults to paid, not approved.

How we built it

Commit Protocol is a hybrid on-chain and off-chain system designed for correctness, not convenience.

On-chain (Solidity and Foundry)

Custom escrow contract on Ethereum using MNEE (ERC-20).
Discord server level treasury accounting.
Commitment state machine: $ \text{FUNDED} \rightarrow \text{SUBMITTED} \rightarrow \text{SETTLED} \text{ or } \text{DISPUTED} $
Batch settlement to push payments automatically.
Secure relayer pattern where only the bot wallet can call sensitive functions.
All balance and settlement invariants enforced on-chain.

Off-chain coordination and agents

Discord bot as the coordination layer where work already happens.
Role-based authorization verified via Discord, not stored on-chain.
AI agents analyze submitted evidence against IPFS-stored specifications.
Agents output a confidence signal used to price disputes.
Cron-based settlement loop that sweeps matured commitments.settlement loop that sweeps matured commitments.

Economic Core

The heart of the protocol is the Dynamic Stake Formula:

$$S_{required} = S_{base} \times M_{time} \times M_{reputation} \times M_{AI}$$

Where:

Time $ (M_{time}) $: Increases the cost of last-minute interference.
Reputation $ (M_{reputation}) $: Protects proven contributors.
AI $ (M_{AI}) $: AI confidence spikes the cost of disputing correct work.

Result: Bad faith becomes economically irrational.

Challenges we ran into

Redesigning the role of AI: Most systems let AI decide who gets paid, which introduces oracle risk. Our AI never decides outcomes; it only modulates economic friction.
Making settlement reliable: Manual claim systems fail when users forget. We built a push-based settlement system using batch execution.
Coordinating without wallets: A secure relayer abstracts wallets away from users while keeping all accounting on-chain.
Defending against malicious contributors: Disputing bad work is cheap; disputing good work is expensive. We rely on incentive compatibility, not approval.