Agreements in Plain English
Describe what you need done, lock up funds, and pick an arbiter to judge whether the work was delivered. Settlement happens on-chain — no middlemen, no manual payouts.
Go to github
AI as the Arbitration Layer
Because your conditions are written in natural language, an AI model can act as a general-purpose judge — reading evidence against what you asked for and deciding whether it was fulfilled. No bespoke evaluation logic needed.
Use Cases
If you can describe it, you can put money behind it.
- Git commit trading — lock up tokens against verifiable repo activity
- Token swaps with subjective or conditional criteria
- Agent-to-agent commitments with human-readable conditions
- Freelance payments that release when work is accepted
- Prediction markets where resolution criteria are stated in prose
- Any bespoke exchange you can describe in plain language
How It Works
Write your conditions in plain English
Lock up ERC20 tokens as payment
The other party submits evidence of fulfillment
An AI oracle reads the evidence against your conditions and releases or rejects the funds
You write the conditions, your counterparty does the work and submits proof. Your chosen oracle evaluates the evidence and settles the funds automatically. The oracle is pluggable — we provide a publicly available, self-deployable AI judge, but a human arbiter, DAO vote, or custom contract can fill the same role.