Autonomous AI agents for prediction markets on Arc.
BRUH is an autonomous prediction-market infrastructure that lets users deploy AI agents with programmable capital and risk constraints, while giving developers a way to build and publish their own forecasting strategies.
Agents generate research, probability estimates, confidence scores, and trade recommendations. BRUH remains responsible for policy enforcement, position sizing, wallet access, scheduling, execution, and accounting.
Developers provide intelligence. Users provide capital and constraints. BRUH provides orchestration and execution.
BRUH turns AI forecasting strategies into deployable economic agents.
A user can:
- Choose a built-in or custom agent
- Select a prediction market or market scope
- Define a deployment mandate
- Set budget, max trade size, exposure, edge, confidence, and risk limits
- Run the agent autonomously on a schedule
- Monitor research, decisions, trades, PnL, and transaction receipts
BRUH evaluates every agent decision before execution. Remote/custom agents never receive direct access to a user's wallet.
Market
↓
Agent Research
↓
Probability Estimate
↓
BUY YES / BUY NO / PASS
↓
BRUH Deployment Policy
↓
Position Sizing + Risk Checks
↓
Execution Queue
↓
Circle Wallet
↓
USDC Trade on Arc
↓
Execution Receipt + Accounting
Focuses on fresh information, news momentum, and recent developments.
Focuses on base rates, historical priors, and conservative probability estimates.
Runs multiple profiles independently and combines their outputs into a consensus decision before execution.
Instead of manually approving every trade, users define a mandate.
A deployment can include:
- Total capital budget
- Maximum position size
- Maximum market exposure
- Minimum edge
- Minimum confidence
- Risk level
- Market scope
- Automatic research
- Automatic execution
- Scheduler interval
BRUH continuously checks eligible markets and only executes decisions that satisfy the deployment policy.
The scheduler includes safeguards for overlapping runs, pause/resume state, maximum markets and trades per cycle, capital already deployed, wallet balance, and existing exposure.
Developers can build their own forecasting strategies with:
npm install @bruhmarket/agent-sdkA custom BRUH agent runs as a remote HTTPS service and implements:
GET /v1/health
POST /v1/run
The agent can return research, probability estimates, confidence, reasoning, key factors, risks, and a recommended action.
The custom agent provides the intelligence only. BRUH still controls wallet custody, budget enforcement, risk checks, position sizing, scheduling, trade authorization, and execution.
A reference implementation is available in:
bruh-agent-consumer/
Custom agents can be registered, versioned, published, and discovered through the BRUH Agent Marketplace.
Build Agent
↓
Host Agent
↓
Register
↓
Verify
↓
Publish
↓
Agent Marketplace
↓
Install
↓
Configure Mandate
↓
Deploy
This allows developers to publish reusable forecasting strategies while users retain control over their own capital and risk configuration.
BRUH uses binary prediction-market contracts on Arc.
Markets support:
- YES / NO outcomes
- USDC collateral
- Buying and selling outcome shares
- Constant-product pricing
- Market closing
- Resolution
- Redemption after settlement
Contracts are located in:
bruh-contracts/
Closed markets are resolved through the OracleJob contract.
Market closes
↓
Resolution job created
↓
Authorised oracle accepts
↓
Oracle posts bond
↓
YES / NO / INVALID submitted with evidence
↓
Dispute window
↓
Finalisation
↓
Market resolved onchain
The oracle job system includes a resolution fee, oracle bond, evidence URI, dispute window, owner arbitration, oracle payout, and YES / NO / INVALID outcomes.
BRUH integrates Circle Developer-Controlled Wallets for agent execution.
For built-in deployments, the live execution path is:
AI Decision
↓
BRUH Policy
↓
Execution Queue
↓
Circle Wallet
↓
Arc
↓
USDC Trade
↓
Transaction Hash
Execution receipts and capital usage are persisted so agent activity can be monitored from the dashboard.
┌──────────────────────┐
│ BRUH UI │
│ Next.js │
└──────────┬───────────┘
│
▼
┌──────────────────────┐
│ BRUH Backend │
│ NestJS │
└──────────┬───────────┘
│
┌──────────────────┼──────────────────┐
│ │ │
▼ ▼ ▼
Built-in Agents Custom Agents Deployment Engine
│ │ │
└──────────────┬───┘ │
▼ ▼
Decision Runtime Policy Engine
│ │
└──────────┬───────────┘
▼
Execution Queue
│
▼
Circle Wallets
│
▼
Arc Testnet
│
▼
Prediction Markets
bruh/
├── bruh-ui/ # Next.js frontend
├── bruh-backend/ # NestJS API, runtime, deployments, execution
├── bruh-contracts/ # Arc prediction-market + oracle contracts
├── bruh-agent/ # @bruhmarket/agent-sdk
├── bruh-agent-consumer/ # Reference custom-agent implementation
├── bruhmarket-sdk-live-test/
├── BRUH_ARCHITECTURE.md
└── README.md
| Layer | Technology |
|---|---|
| Frontend | Next.js, React, Tailwind CSS |
| Backend | NestJS, TypeScript |
| Database | Supabase |
| Agent Runtime | @bruhmarket/agent-sdk |
| LLM | Anthropic |
| Wallets | Circle Developer-Controlled Wallets |
| Chain | Arc Testnet |
| Contracts | Solidity, Foundry |
| RPC / EVM | viem |
| Settlement Asset | USDC |
| Hosting | Vercel + Railway |
BRUH separates agent intelligence from execution authority.
A third-party agent cannot directly access wallet credentials, execute arbitrary transactions, ignore user budgets, bypass risk limits, or change deployment policy.
Developer Agent
↓
Protocol Validation
↓
BRUH Runtime
↓
Deployment Policy
↓
Execution Queue
↓
Circle Wallet
↓
Arc
BRUH includes safeguards for autonomous operation:
- Scheduler locking
- Pause / resume controls
- Run persistence
- Execution receipts
- Capital accounting
- Market exposure checks
- RPC retry / backoff
- Short-lived market-state caching
- Reduced scheduler pressure
- Maximum work limits per deployment cycle
Implemented:
- Built-in forecasting agents
- Shared agent runtime
- Ensemble / consensus decisions
- Deployment mandates
- Autonomous scheduler
- Policy enforcement
- Position sizing
- Circle wallet execution
- Arc USDC trades
- Execution receipts
- Capital accounting
- Agent SDK
- Custom-agent protocol
- Agent registry
- Agent versions
- Agent installation
- Agent Marketplace
- Market contracts
- Oracle-based market resolution
- Dashboard and deployment UI
Built-in deployments support the full live execution flow.
Custom agents can currently be built, hosted, registered, verified, installed, invoked, and evaluated through BRUH's policy layer. Direct live execution for custom deployments remains intentionally gated in the current beta.
AI agents are becoming capable of generating useful forecasts, but intelligence alone is not enough.
BRUH connects agent intelligence to programmable money, explicit risk limits, autonomous execution, and verifiable onchain outcomes.
The goal is to create an open ecosystem where developers can publish forecasting strategies, users can deploy them with their own mandates, and BRUH provides the trusted infrastructure connecting intelligence to capital.
Programmable Money Hackathon — Build on Arc
- Arc Testnet
- USDC
- Circle Wallet Infrastructure
- Autonomous AI Agents
- Prediction Markets
Hackathon / experimental software. Review contracts and infrastructure carefully before using with real funds.