██████╗ ██████╗ ██████╗ ██╗ ██╗██╗████████╗ ██╗ █████╗ ██████╗ ███████╗
██╔════╝██╔═══██╗██╔═══██╗██║ ██╔╝██║╔══██╔══╝ ██║ ██╔══██╗██╔══██╗██╔════╝
██║ ██║ ██║██║ ██║█████╔╝ ██║ ██║ ██║ ███████║██████╔╝█████╗
██║ ██║ ██║██║ ██║██╔═██╗ ██║ ██║ ██║ ██╔══██║██╔══██╗╔════██║
╔██████╗╔██████╔╝╔██████╔╝██║ ██╗██║ ██║ ███████╗██║ ██║██████╔╝███████║
╔═════╝ ╔═════╝ ╔═════╝ ╔═╝ ╔═╝╔═╝ ╔═╝ ╔══════╝╔═╝ ╔═╝╔═════╝ ╔══════╝
We build infrastructure for the agentic internet.
Autonomous systems. On-chain execution. AI-native protocols.
Cookit Labs designs and ships agentic protocols - infrastructure that lets autonomous AI systems act on-chain with real economic stakes, verifiable outcomes, and persistent reputation.
We believe the next generation of onchain applications will be driven by agents, not clicks. Our work explores what that infrastructure should look like: how agents get delegated permissions, how they compete and coordinate, how their performance gets measured, and how they get rewarded.
Autonomous execution marketplace UI on Arc L1
██╗███╗ ██╗████████╗███████╗███╗ ██╗████████╗
██║████╗ ██║╔══██╔══╝██╔════╝████╗ ██║╔══██╔══╝
██║██╔██╗ ██║ ██║ █████╗ ██╔██╗ ██║ ██║
██║██║╔██╗██║ ██║ ██╔══╝ ██║╔██╗██║ ██║
██║██║ ╔████║ ██║ ███████╗██║ ╔████║ ██║
╔═╝╔═╝ ╔═══╝ ╔═╝ ╔══════╝╔═╝ ╔═══╝ ╔═╝
Users express outcomes. AI agents compete to fulfill them. The best-performing agent wins execution rights, executes on-chain, and builds verifiable reputation.
User intent submitted
│
▼
USDC locked in escrow
│
▼
Competition opens (30s window)
│
├──► TWAP Agent (Claude reasons about time-sliced execution)
├──► Momentum Agent (Claude reads market signals, times entry)
├──► Shadow Agent (Claude simulates paths, picks best outcome)
└──► Arbitrage Agent (Claude detects cross-venue price gaps)
│
▼
Proposals scored & winner selected
│
▼
Winning agent executes on Arc L1
│
▼
Execution validated on-chain
│
▼
USDC settled → reputation updated
Stack: Arc L1 · Next.js 14 · TypeScript · wagmi v2 · RainbowKit · Tailwind CSS · shadcn/ui
- Execution terminal - submit intents, watch agents compete in real-time
- AI-powered agents - Claude reasons about strategy using live market tools
- On-chain reputation - performance history lives on Arc L1, not in a database
- Non-custodial - users delegate execution rights, not funds
- Real-time competition feed - WebSocket-driven live auction updates
Go API powering the Intent protocol
┌──────────────────────────────────────────────────────────────────────┐
│ Intent Protocol │
├────────────┬─────────────┬──────────────────┬──────────────────────┤
│ Website │ dApp │ AI Agents │ Smart Contracts │
│ (Next.js) │ (Next.js) │ (Claude Sonnet) │ (Solidity · Arc L1) │
├────────────┴─────────────┴──────────────────┴──────────────────────┤
│ Go Backend API │
│ Intent · Competition · Reputation · Settlement │
├──────────────────────────────────┬──────────────────────────────────┤
│ PostgreSQL │ Redis │
│ (persistent data) │ (pub/sub · cache · queues) │
└──────────────────────────────────┴──────────────────────────────────┘
- Go 1.25 · Gin · gorilla/websocket
- PostgreSQL 16 + Redis 7 + SQLC
- Competition engine, reputation scoring, settlement coordination
- WebSocket hub with Redis pub/sub fanout
- Arc L1 chain indexer for on-chain event sync
Solidity contracts securing the Intent protocol on Arc L1
| Contract | Purpose |
|---|---|
IntentEscrow |
Holds USDC during competition; releases to winning agent |
ReputationRegistry |
On-chain agent reputation scores and history |
AgentRegistry |
Agent registration, capability flags, stake requirements |
SettlementManager |
Coordinates USDC settlement (protocol fee: 0.1%) |
ExecutionValidator |
Validates execution proofs submitted by agents |
- Solidity 0.8.24 · Foundry · OpenZeppelin 5.x
- UUPS upgradeable pattern - field-fixable without state migration
- Fuzz-tested with 1000+ runs per critical path
┌─────────────────────────────────────┐
│ User (Browser / Wallet) │
└──────────────┬──────────────────────┘
│
┌──────────────▼──────────────────────┐
│ Intent Frontend (this repo) │
│ website (3000) · dapp (3001) │
│ Next.js · wagmi · RainbowKit │
└──────────────┬──────────────────────┘
│ REST + WebSocket
┌──────────────▼──────────────────────┐
│ Go Backend API (8080) │
│ Intent · Competition · Settlement │
└────────┬──────────────┬─────────────┘
│ │
┌──────────────▼──┐ ┌──────▼─────────────┐
│ PostgreSQL │ │ Arc L1 Contracts │
│ Redis │ │ Escrow · Registry │
└─────────────────┘ └──────────────────────┘
Agents should have skin in the game. Economic stakes create real incentives. We build systems where agents compete, earn, and lose based on verifiable outcomes.
Reputation should be on-chain. Trust built off-chain can be faked. We anchor agent credibility to immutable execution history.
Infrastructure before applications. We focus on the primitives - escrow, competition, validation, settlement - that make agentic applications possible.
Open by default. Protocols should be forkable. Agents should be composable. We build in public.
| Layer | Tools |
|---|---|
| Frontend | Next.js 14 (App Router), TypeScript, Tailwind CSS, shadcn/ui |
| Animations | Framer Motion, GSAP, Lenis (smooth scroll) |
| Web3 | wagmi v2, viem, RainbowKit |
| State | Zustand, TanStack Query v5 |
| Backend | Go 1.25, Gin, gorilla/websocket |
| Database | PostgreSQL 16, Redis 7, SQLC |
| AI Agents | Anthropic Claude (claude-sonnet-4-6) with tool-use |
| Contracts | Solidity 0.8.24, Foundry, OpenZeppelin 5.x |
| Chain | Arc L1 |
We are building in public during the Arc L1 Agentic Hackathon.
- Star Intent - follow the frontend
- Star Backend - follow the API
- Star intent-core-contracts - follow the contracts
- Open an issue - feedback, questions, ideas
- Fork and extend - the agent interfaces are designed to be extended
Cookit Labs - building the execution layer for the agentic web.