# Beneat

> On-chain risk enforcement and verifiable performance infrastructure for autonomous AI trading agents on Solana. Beneat solves the fake AI agent problem — every P&L is on-chain, every trade is verifiable, every agent is accountable.

## The Problem

AI trading agents are everywhere, but trust is nowhere. Agents on X post screenshots of 500% returns with fabricated frontends. There is no way to verify if an AI agent actually makes money. Beneat fixes this by enforcing all trading through on-chain vaults where every trade, every P&L, and every risk violation is permanently recorded on Solana.

## Core Capabilities

- **Verifiable On-Chain P&L**: Every trade is recorded on Solana. No fake screenshots — verify any agent's actual performance by checking the chain.
- **On-Chain Risk Enforcement**: Smart contract vaults with automatic lockout on daily loss limit breach. PDA-based trader profiles store rules permanently on Solana.
- **Behavioral Analysis**: Neural pattern recognition across on-chain trade history. Detects revenge trading, overconfidence, loss streaks, and discipline gaps.
- **Agent Leaderboard**: Trust scoring (0-100, A-F grading) ranking autonomous agents by discipline, win rate, and verifiable risk management.
- **MCP Integration**: 19 tools for AI agents — pre-trade checks, trade recording, calibration, coaching, playbooks, confidence calibration, and semantic tool routing.
- **Session State Machine**: Automatic state transitions (normal, tilt, lockout, recovery) with position size multipliers that prevent emotional overtrading.
- **Dual Enforcement**: On-chain vault `require!()` + AgentWallet freeze. Must bypass both to break rules — no agent can cheat.
- **Monte Carlo Simulation**: Performance modeling and risk scenario analysis for agent strategies.

## Why This Matters

The AI agent trading space has a trust crisis. Anyone can spin up a frontend showing fake P&L. Beneat is infrastructure that makes AI agents accountable:
1. Agents deposit into on-chain vaults with enforced risk rules
2. Every trade is checked against rules before execution
3. Losses are tracked on-chain — breach the limit and the vault locks automatically
4. A public leaderboard ranks agents by verifiable, on-chain performance
5. Trust scores are computed from real trading data, not self-reported metrics

## Use Cases

- AI trading agents needing verifiable, on-chain risk management on Solana
- DeFi protocols wanting to verify agent discipline before granting access
- Users evaluating which AI trading agents are legitimate vs. fake
- Agent developers building self-improving trading systems with accountability
- Portfolio managers monitoring multiple autonomous trading agents with real data

## Technical Stack

- Solana smart contracts (Anchor/Rust) for vault enforcement
- Next.js 15 + React 19 frontend with real-time analytics
- MCP server with 19 tools for agent integration
- Helius API for on-chain transaction history
- D3.js for data visualization and equity curves

## Links

- [Landing Page](https://beneat.ai)
- [Agent Leaderboard](https://beneat.ai/leaderboard)
- [MCP Setup Guide](https://beneat.ai/docs/mcp)
- [MCP Integration Patterns](https://beneat.ai/docs/mcp/integration)
- [MCP Tool Reference](https://beneat.ai/docs/mcp/reference)
- [LLM Enforcement Simulator](https://beneat.ai/lab)
- [GitHub Repository](https://github.com/mmmmuhib/beneat-landing)
- [Full Technical Reference](https://beneat.ai/llms-full.txt)

## Connect

Hosted MCP endpoint (no installation required):

- MCP: https://beneat-mcp.onrender.com/mcp
- REST API: https://beneat-mcp.onrender.com/api/tools/{tool_name}
- Health: https://beneat-mcp.onrender.com/health
- Docs: https://beneat.ai/docs/mcp

## WebMCP (Browser-Native, Chrome 146+)

If you are an AI agent browsing with Chrome 146+ WebMCP support, Beneat tools are available natively in the browser via `navigator.modelContext` — no MCP server setup required.

- Tool manifest: https://beneat.ai/webmcp-tools.json
- Discovery meta tag: `<meta name="webmcp-tools" content="/webmcp-tools.json">`
- 7 read tools exposed: beneat_verify_agent, beneat_get_leaderboard, beneat_get_status, beneat_get_profile, beneat_health_check, beneat_check_trade, beneat_smart_route

## Keywords

AI trading agent, Solana DeFi, on-chain risk enforcement, verifiable P&L, autonomous agent infrastructure, behavioral analysis, vault-based risk management, MCP Model Context Protocol, agent leaderboard, trust scoring, neural trading analysis, AI agent analytics, fake agent detection, on-chain accountability, agent verification