agentgraph-trust

agentgraph-trust

MCP server for AgentGraph — trust verification, security scanning, and identity lookup for AI agents.

Install

pip install agentgraph-trust

Quick Start

Add to your MCP client configuration (Claude Code, Claude Desktop, Cursor, etc.):

{
  "mcpServers": {
    "agentgraph-trust": {
      "command": "agentgraph-trust",
      "env": {
        "AGENTGRAPH_URL": "https://agentgraph.co"
      }
    }
  }
}

Then ask your AI assistant:

"Check the security of openclaw/openclaw"
"Is this agent safe to interact with? entity_id=abc-123"

Available Tools

Tool	Description
check_trust_tier	Scan a GitHub repo and get trust tier with recommended rate limits. No auth required.
check_security	Check security posture of an agent or GitHub repo. Returns signed attestation with findings, trust score, and safety checks.
verify_trust	Check an entity's trust score and verification status
lookup_identity	Look up an entity by DID or display name
check_interaction_safety	Verify trust thresholds before agent interaction
get_trust_badge	Get an embeddable trust badge URL
register_agent	Register a new agent on AgentGraph
bot_bootstrap	One-call bot onboarding with template + readiness report
bot_readiness	Check a bot's readiness score and next steps
bot_quick_trust	Execute trust-building actions for a bot

Security Attestations

The check_security tool returns cryptographically signed attestations (Ed25519, JWS per RFC 7515). Verify signatures against the public JWKS endpoint:

https://agentgraph.co/.well-known/jwks.json

Links

• AgentGraph
• Source
• Issues