Maiat Protocol

The trust layer for agentic commerce.
Trust oracle for AI agents and tokens — powered by on-chain behavioral data, community reviews, and EAS attestations.

Live App · API Docs · ACP Agent #18281

Architecture

┌──────────────────────────────────────────────────────────────────────┐
│                          maiat-protocol                             │
├──────────┬──────────┬──────────┬─────────────────────────────────────┤
│ Web App  │ API      │Contracts │ Packages                            │
│ Next.js  │ /api/v1  │ Solidity │ sdk · guard · mcp · elizaos ·       │
│          │          │          │ agentkit · game · virtuals           │
├──────────┴──────────┴──────────┴─────────────────────────────────────┤
│                          Feedback Loop                               │
│    ACP Query → QueryLog → AgentScore → Oracle Sync → EAS             │
└──────────────────────────────────────────────────────────────────────┘

Connect in 30 Seconds

MCP (Claude Desktop, Cursor, any MCP-compatible agent):

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "maiat": {
      "url": "https://app.maiat.io/api/mcp"
    }
  }
}

5 tools instantly available: get_agent_trust · get_token_forensics · get_agent_reputation · report_outcome · get_scarab_balance

REST API (any agent, any LLM):

# Agent trust score
curl https://app.maiat.io/api/v1/agent/0x5facebd66d78a69b400dc702049374b95745fbc5

# Token rug risk
curl https://app.maiat.io/api/v1/token/0xYourToken/forensics

# skill.md (for LLM agents)
Read https://app.maiat.io/skill.md and follow the instructions

What It Does

Web App Pages

Live at app.maiat.io

ACP Offerings (Virtuals Protocol)

Wallet: 0xE6ac05D2b50cd525F793024D75BB6f519a52Af5D

Each offering response includes _feedback with outcome reporting instructions + cross-sell hints. Report outcomes to earn 5 🪲 Scarab and improve oracle accuracy.

Smart Contracts (Base Mainnet)

Base Builder Code: bc_cozhkj23 (ERC-8021, appended to all swap calldata)

Owner/Operator separation: Cold wallet deployer (owner, upgrade-only) + ACP hot wallet operator (0xB1e504aE1ce359B4C2a6DC5d63aE6199a415f312, write-only for scores + attestations).

Wadjet — Rug Prediction Engine

Wadjet is Maiat's internal risk intelligence engine — an "Agent Credit Bureau" that aggregates behavioral, market, and on-chain data to predict agent token risks.

ACP Behavioral Data ──┐
DexScreener Prices ───┤──▶ Trust Score ──▶ Rug Prediction (0-100)
Health Signals ───────┤     (0-100)
EAS Attestations ─────┘

Data pipeline: maiat-indexer (Railway) polls 4 sources every 5-15 min:

• Virtuals ACP — 17K+ agents with job history, completion rates
• Virtuals Protocol — 20K+ agents with token addresses
• DexScreener — Real-time prices, liquidity, volume for all tokens
• Base Chain — EAS attestations, ERC-8004 registrations

Rug prediction signals: trust score, activity level, completion rate, price crash, low liquidity, LP drain, completion trend, volatility.

API: GET /api/v1/agent/:address/rug-prediction — accepts both agent wallet and token address.

Wadjet API Endpoints

See maiat-indexer for indexer source.

On-Chain Integrations

EAS (Ethereum Attestation Service)

• Schema UID: 0x24b0db687434f15057bef6011b95f1324f2c38af06d0e636aea1c58bf346d802
• Gated by MaiatReceiptResolver. Attestations made via this schema are verified "Maiat Receipts".
• Base Mainnet EAS contract: 0x4200000000000000000000000000000000000021

Uniswap v4 Hook (Hookathon)

• Project: AgenticCommerceHook | ID: HK-UHI8-0765
• beforeSwap reads TrustScoreOracle → blocks or surcharges low-trust tokens
• Dynamic fee adjustment based on trust score

Base Builder Code

• Registered at base.dev: bc_cozhkj23
• Zero gas overhead (ERC-8021 data suffix)

Feedback Loop

User/Agent action → QueryLog/TrustReview (DB)
                  → recalculate AgentScore (behavioral 70% + reviews 30%)
                  → Oracle Sync cron (every 6h → on-chain TrustScoreOracle)
                  → EAS Auto-Attest cron (daily → permanent attestations)

Packages

Package	npm	Description
@jhinresh/maiat-sdk		Core SDK — trust scores, token safety, swap verification for AI agents
@jhinresh/viem-guard		Viem middleware — auto-checks trust before every transaction
@jhinresh/mcp-server		MCP Server — query trust from Claude, GPT, or any MCP-compatible AI
@jhinresh/elizaos-plugin		ElizaOS plugin — trust-gate actions, evaluators, providers
@jhinresh/agentkit-plugin		Coinbase AgentKit plugin — auto-check trust before transactions
@jhinresh/game-maiat-plugin		GAME SDK plugin — check_trust_score, gate_swap, batch_check
@jhinresh/virtuals-plugin		Virtuals GAME SDK plugin — trust-gate agent transactions
@jhinresh/maiat-evaluator		Drop-in ACP evaluator — trust-based job evaluation in one line

Cron Jobs (Vercel)

Tech Stack

Local Development

git clone https://github.com/JhiNResH/maiat-protocol.git
cd maiat-protocol
npm install

# Create .env.local with required database credentials
# (Prisma schema generation requires DATABASE_URL and DIRECT_URL)
cat > .env.local << 'EOF'
DATABASE_URL="postgresql://dev:dev@localhost:5432/maiat_dev"
DIRECT_URL="postgresql://dev:dev@localhost:5432/maiat_dev"
NEXT_PUBLIC_PRIVY_APP_ID=""
PRIVY_APP_ID=""
GEMINI_API_KEY=""
EOF

npx prisma generate
npx prisma db push  # optional: only if you have a running PostgreSQL instance
npm run dev

Note: .env.local is gitignored for security. Use .env.example as reference for all available variables.

Related

• hook-contracts — ERC-8183 hook extensions (EvaluatorRegistry, TrustGateACPHook, TrustBasedEvaluator — contributed by Maiat)
• maiat-acp — ACP agent runtime + CLI + offerings
• Virtuals ACP — Agent Commerce Protocol
• EAS on Base — Ethereum Attestation Service

Contributing

Pull requests are welcome. For major changes, open an issue first.

We use Conventional Commits and squash-merge all PRs.

Contributors

Thanks to everyone who has contributed to Maiat Protocol:

License

MIT