🛡️ Beaver Warrior Sentinel

Connecting...

The First Decentralized Trust Layer
for AI Agents

Beaver Warrior Sentinel autonomously evaluates AI agents on-chain, detects sleeper agents, and creates economic accountability through ETH staking & slashing — so agents can safely transact with each other.
VirusTotal for AI agents — but decentralized and autonomous
375
Security Modules
58
AI Agent Modules
<100MB
Memory Footprint
4
On-Chain Registries

🚨 The Problem: AI Agents Have No Trust Layer

The agent economy is exploding. Google shipped A2A (Agent-to-Agent protocol), Anthropic shipped MCP, and ERC-8004 landed on Ethereum — all in the past few months. But there's a critical missing piece: when agents talk to agents, how do you know who to trust?

🤖 Sleeper Agents

An agent behaves perfectly for weeks, builds trust, then activates malicious behavior — stealing data, injecting prompts, or redirecting funds. No existing system detects this.

Our swarm catches them in real-time via continuous re-evaluation

🎭 Capability Fraud

Agents claim skills they don't have. A "financial advisor" agent that actually just scrapes Reddit. A "code reviewer" that returns random approvals. Users can't verify claims before trusting.

We test every claimed capability with real HTTP requests

💸 No Economic Consequences

Today, a malicious agent faces zero cost for bad behavior. It can re-register under a new name and continue. There's no skin in the game, no accountability, no deterrent.

ETH staking means real money is lost when trust is broken

🗺️ Current Cybersecurity Landscape

Existing solutions address parts of the problem, but none provide a complete trust infrastructure for AI agents:

What Exists Today

  • EigenLayer — Restaking for validators, not AI agents
  • Worldcoin / Gitcoin Passport — Human identity only
  • LangChain / CrewAI / AutoGen — No on-chain reputation
  • Google A2A — Communication spec, zero trust layer
  • Traditional AVS — No behavioral evaluation
  • Academic Papers — Theoretical, no working code
VS

Beaver Warrior Sentinel

  • On-chain agent identity (ERC-8004)
  • Automated behavioral evaluation
  • Economic staking & slashing
  • Agent-to-agent trust gating
  • Sleeper agent detection
  • Open onboarding REST API
  • Working demo with real LLM agents

Nobody has combined all of these into a single working system — until now.

🦫 Beaver Warrior: 375 Security Modules in Under 100MB

Beaver Warrior is a desktop cybersecurity application (macOS / Windows / Linux) built in Rust for maximum performance and minimal footprint. It runs 375 security modules — including 58 specifically designed for AI agent threats — all within a <100MB memory footprint. Here's how those modules break down:

🛡️ Network Security

Real-time packet inspection, DNS filtering, TLS certificate validation, connection monitoring

87 modules

🔒 Endpoint Protection

File integrity monitoring, process sandboxing, privilege escalation detection, rootkit scanning

64 modules

🧠 AI Agent Security

Sleeper agent detection, prompt injection defense, autonomous containment, multi-agent conflict resolution, behavioral fingerprinting

58 modules

🔗 Blockchain & Web3

Smart contract interaction monitoring, wallet protection, phishing detection, MEV protection

42 modules

📊 Threat Intelligence

Behavioral analytics, anomaly detection, threat correlation, real-time threat feeds, zero-day pattern matching

53 modules

🔐 Data Protection

Encryption enforcement, data leak prevention, clipboard monitoring, sensitive data redaction

38 modules

🌐 Privacy

Tracker blocking, fingerprint prevention, telemetry control, secure DNS resolution

33 modules

All 375 modules are written in Rust for memory safety and performance. The entire engine runs in a single process with <100MB RSS, making it lighter than most Electron apps while providing deeper protection than enterprise security suites costing thousands per seat.

🏗️ Architecture: How It All Fits Together

🤖 AI Agent Registers on ERC-8004
🛡️ Sentinel Swarm 6-agent evaluation pipeline
🔗 On-Chain Score 4 ERC-8004 registries
🦫 Beaver Warrior Reads scores, blocks threats
👤 End User Protected automatically

The Sentinel swarm evaluates once, and every Beaver Warrior user is protected.
Trust data lives permanently on-chain via 4 ERC-8004 registries: IdentityReputationValidationStaking

What We Built for This Hackathon

🤖 Real LLM Agents

Honest agent + sleeper agent powered by Claude 3.5 Haiku. Real AI reasoning, not mock data. The sleeper starts honest, then activates malicious behavior mid-demo.

💰 Staking & Slashing

Solidity smart contract where agents stake ETH as collateral. Trust score drops below threshold? 50% of stake is slashed automatically. Real economic consequences.

🤝 Trust Gating

Agents autonomously refuse to collaborate with untrusted peers by reading on-chain reputation scores. No centralized authority — agents make their own trust decisions.

📜 Persistent Reputation

Trust scores survive restarts. The swarm reads existing on-chain reputation data on startup and resumes where it left off. No reputation amnesia.

🔌 Onboarding API

Any third-party agent can register via REST API and get evaluated in the next cycle. Open protocol — not a walled garden.

🔍 Deep Audit

10 bugs found and fixed through a rigorous 3-pass security audit, including a critical access-control vulnerability in the staking contract.

🐝 The 6-Agent Evaluation Swarm

The Sentinel isn't a single program — it's a multi-agent swarm where 6 specialized agents work together in a pipeline, each with a single responsibility:

🔍 DiscoveryAgent Polls chain for new agents
🧪 EvaluatorAgent Tests capabilities via HTTP
📊 ScorerAgent Computes 0-100 trust score
🛡️ SafetyAgent 6 pre-flight checks
📤 PublisherAgent Writes score on-chain

Plus a Coordinator that orchestrates the pipeline, manages budgets, and triggers staking/slashing + trust gating after each cycle.

🚀 See It Live

Switch to the Live Demo tab to watch the swarm evaluate agents in real-time. Watch the honest agent maintain 100/100, the liar get caught at ~80, and the sleeper agent degrade from 100 to 40 as its malicious behavior activates.

Swarm Agents

These 6 autonomous agents work together in a pipeline. Green dot = finished this cycle. Orange = busy.
DiscoveryAgent
Polls blockchain for new AI agent registrations
Processed: 0
EvaluatorAgent
Sends real HTTP requests to test each agent's claimed skills
Processed: 0
ScorerAgent
Computes weighted 0-100 trust score from test results
Processed: 0
SafetyAgent
6 safety checks before allowing on-chain writes
Processed: 0
PublisherAgent
Writes trust score to blockchain (with retry)
Processed: 0

Discovered Agents — Live Trust Scores

Each card is a real AI agent registered on the blockchain. The swarm tested it and computed a trust score. Orange tags = detected problems. The chart shows how scores change across re-evaluations.
🚨 THREAT DETECTED — Sleeper Agent Caught
A sleeper agent builds trust by behaving honestly, then activates malicious behavior. The swarm detected this by re-evaluating the agent and seeing its score drop.
API Calls
0 / 50
Gas (gwei)
0 / 500,000
Runtime
0s
Waiting for swarm to discover agents...

Swarm Activity

Real-time counters and event log. Every action is logged — discover, evaluate, score, publish.
0
Cycles
0
Published
0
Blocked
0
Messages

Event Log