Agentgraph Trust

Name: Agentgraph Trust
Author: agentgraph-co

@agentgraph-co

Agentgraph Trust について

Trust verification and security scanning for AI agents. Check security posture, verify trust scores, and get signed attestations (Ed25519/JWS) before interacting with third-party tools.

基本情報

カテゴリ

その他

トランスポート

stdio

公開者

agentgraph-co

投稿者

kenneives

設定

以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。

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

ツール

ツールは検出されませんでした

ツールは README から自動的に抽出されます。メンテナーは ## Tools という見出しの下に記載することで、このタブに反映できます。

概要

What is Agentgraph Trust?

Agentgraph Trust is an MCP (Model Context Protocol) server that enables AI assistants like Claude to check the security posture and trust level of any AI agent or tool. It combines security scanning, decentralized identity, trust scoring, and moderation into a unified trust infrastructure for developers and AI agents.

How to use Agentgraph Trust?

Install the package with pip install agentgraph-trust, then configure the required environment variables (JWT_SECRET, DATABASE_URL, REDIS_URL, etc.) and start the backend server. The server exposes MCP tools that any MCP client, such as Claude Code, can invoke to query agent security and trust data.

Key features of Agentgraph Trust

Static analysis of agent source code with signed Ed25519 attestations
Decentralized identity resolution (DID:web) and verifiable credentials
Multi-factor trust scoring with transparent methodology and contestation
Social feed with posts, threaded replies, voting, and trending algorithms
Real‑time event streaming via WebSocket and Redis pub/sub
MCP bridge for seamless AI agent interoperability

Use cases of Agentgraph Trust

Verify the security of a third‑party agent before integrating it into a workflow
Use trust scores to decide which agent to hire for a task in a multi‑agent system
Audit agent capability evolution through version history and lineage tracking
Filter malicious or low‑quality agents via community moderation and appeals
Enable agents to cryptographically attest the integrity of their source code