OneSet

OneSet is a session-based trading and prediction application that enables instant, gasless user and AI-agent interactions off-chain, with secure on-chain settlement only once at the end of a session. Built using Yellow Network, Uniswap v4, and LI.FI, OneSet delivers Web2-like speed while preserving Web3 security and verifiability.

Problem Modern on-chain trading and prediction apps suffer from fundamental UX and cost issues:

• Every interaction requires a blockchain transaction
• High gas fees make micro-transactions impractical
• Latency breaks real-time trading and agent-driven strategies
• AI agents cannot safely execute frequent actions on-chain As a result, many high-frequency or session-based financial interactions are either impossible or economically inefficient on-chain.

Solution OneSet introduces a session-based model for DeFi interactions:

• Users deposit funds once
• Perform unlimited instant actions off-chain (trades, predictions, agent execution)
• Settle the net result on-chain only once when the session ends This approach dramatically reduces gas costs, improves UX, and enables entirely new classes of applications such as AI-driven trading and micro-markets.

So for the frontend we use Next Js with TypeScript, the Smart Contracts are deployed on Arbitrum.

We use Yellow’s state channels and session keys to:

• Enable real-time, gasless interactions
• Support session-based spending limits
• Secure off-chain balances with cryptographic guarantees
• Compress dozens of actions into a single on-chain settlement

Uniswap v4 is used for final execution and price discovery, not for every interaction. Trades are simulated and netted off-chain during a session Agentic Finance

• AI agents make decisions off-chain in real time Hooks are used to enforce:
• Slippage limits
• Risk constraints
• Execution conditions
• Execution remains transparent, verifiable, and safe

Using LI.FI, we enable:

• Deposits from any EVM chain
• Swap + bridge + deposit in a single flow
• Seamless capital movement for users and AI agents