TrstLyr Protocol

Before your agent trusts another agent with money, code, or data — it checks TrstLyr.

TrstLyr is the trust layer for the agent internet. It aggregates signals from GitHub, ERC-8004, Twitter/X, Self Protocol ZK proof-of-human, ClawHub, Moltbook, and post-interaction behavioral attestations into verifiable trust scores — anchored on-chain via EAS on Base. Open source, Apache 2.0, free to query.

Try it now

curl https://api.trstlyr.ai/v1/trust/score/github:yourusername

{
  "subject": "github:yourusername",
  "trust_score": 50.1,
  "confidence": 0.72,
  "risk_level": "medium",
  "recommendation": "review",
  "signals": [ "..." ]
}

No API key. No wallet. No signup.

Why it matters

• Agents are making real decisions with real money. A credential stealer disguised as a weather skill sat on ClawHub for weeks before accidental discovery. There is no trust infrastructure — until now.
• No single signal is enough. A fresh GitHub account can have an on-chain ERC-8004 token. A high-karma Moltbook agent can be a Sybil. TrstLyr fuses signals from 7 providers and makes honest behavior the mathematically dominant strategy (Ev-Trust, arXiv:2512.16167).
• Trust should be infrastructure, not a product. Like DNS or SSL, trust is a public good. Free core API. Self-hostable. Apache 2.0 forever.

Quick start

REST API

# Trust score (free, no key)
curl https://api.trstlyr.ai/v1/trust/score/github:tankcdr

# Pre-action trust gate — should my agent proceed?
curl -X POST https://api.trstlyr.ai/v1/trust/gate \
  -H "Content-Type: application/json" \
  -d '{"counterparty":"github:tankcdr","action":"delegate"}'

# Anchor score on-chain (1st free, then $0.01 USDC via x402)
curl -X POST https://api.trstlyr.ai/v1/attest \
  -H "Content-Type: application/json" \
  -d '{"subject":"github:tankcdr"}'

MCP (Claude Desktop / any MCP runtime)

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

{
  "mcpServers": {
    "trstlyr": {
      "command": "npx",
      "args": ["-y", "@trstlyr/mcp"]
    }
  }
}

Gives your agent four tools: trust_query, should_proceed, trust_explain, trust_batch.

Providers

All providers run in parallel. One query fans out to every applicable provider.

Behavioral attestations

After two agents complete a transaction, delegation, or collaboration, either party can submit a behavioral attestation — anchored on-chain as an EAS attestation on Base Mainnet.

# Submit a post-interaction attestation
curl -X POST https://api.trstlyr.ai/v1/attest/behavioral \
  -H "Content-Type: application/json" \
  -d '{
    "subject": "github:counterparty",
    "interaction_type": "task_delegation",
    "outcome": "completed",
    "rating": 5,
    "evidence_uri": "https://..."
  }'

# Query behavioral history for any subject
curl https://api.trstlyr.ai/v1/trust/behavior/github:tankcdr

Static signals (GitHub age, ERC-8004 registration) describe what an agent claims. Behavioral attestations prove what it does. Behavioral EAS schema: 0xcc161b722fd96c32cb8515fe86c6c76c23aefddbf3eb9973dadc37298d1c65b9

On-chain anchoring

Trust scores are anchored as EAS attestations on Base Mainnet.

How it works: Call POST /v1/attest with a subject. TrstLyr computes the trust score, serializes it (subject, score, confidence, risk level, signal summary), and submits an on-chain EAS attestation. The attestation UID is returned immediately. After the first free attestation, subsequent calls return HTTP 402 with x402 payment instructions — your agent pays $0.01 USDC and the attestation is created automatically.

Architecture

Agent / CLI / Platform
        │
        ▼
   REST API (Fastify)           MCP Server
        │                           │
        ▼                           ▼
   Trust Aggregation Engine
   ├── Identity Resolver (cross-namespace linking)
   ├── Signal Dispatcher (parallel fan-out, 10s timeout)
   ├── Scoring Engine (Subjective Logic + Ev-Trust)
   └── Cache (TTL per signal type)
        │
        ▼
   7 Signal Providers ──► EAS Attestation Bridge (Base L2)

Self-host: git clone https://github.com/trstlyr/trstlyr-protocol && docker compose up -d

Full details: Architecture | Specification | Provider Guide

Contributing

Apache 2.0 — PRs welcome.

Add a new provider:

1. Implement the Provider interface (docs)
2. Register it in the provider index
3. Open a PR with tests and example subjects

Issues, edge cases, signal provider proposals: GitHub Issues

Discussion: OpenClaw Discord

Status

Live. Deployed on Railway. Serving trust queries now.

Built by Chris Madison · Powered by Charon