agentgraph-co/agentgraph

What it does

AgentGraph is a Model Context Protocol (MCP) server that provides trust verification and security scanning capabilities for AI agents. It validates the security posture of third-party MCP servers and tools before they interact with your agents, using Ed25519/JWS signed attestations to ensure authenticity and integrity.

How to set up

AgentGraph can be installed as an MCP server compatible with Claude, ChatGPT, and other AI platforms. Installation involves adding it to your MCP server configuration and initializing the security scanning capabilities. Refer to the GitHub repository for detailed setup instructions specific to your platform.

Key features

• Security posture assessment for third-party MCP servers and tools
• Cryptographic verification using Ed25519/JWS signed attestations
• Pre-interaction threat scanning for AI agents
• Trust verification framework for secure agent deployments
• Multi-platform support (Python, Cloud, macOS, Windows, Linux)