agentidentityprotocol

Overview

Agent Identity Protocol (AIP) is an open-source standard for authentication, attestation, authorization, and governance of AI agents. It serves as a policy enforcement proxy for MCP, implementing a zero-trust security model where agents are issued cryptographically-signed authentication tokens (AATs) that are verified before any tool execution. The system includes DLP scanning, human-in-the-loop approval capabilities, and comprehensive audit logging to meet compliance requirements like SOC 2 and GDPR.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• read_file

  Read the contents of a file

  Try:Read the contents of the report.pdfTry:Show me the configuration file
• list_directory

  List files and directories in a specified path

  Try:List all files in the documents folderTry:Show what's in the current directory
• git_status

  Check the current status of a git repository

  Try:Check git status for the current projectTry:Show me any uncommitted changes
• write_file

  Write content to a file

  Try:Save this analysis to results.txtTry:Create a new configuration file with these settings
• exec_command

  Execute a system command

  Try:Run the database backup scriptTry:Execute the system update command

Note: Tool names inferred from examples in the README and policy configuration, not from an explicit tool listing section

Comparable tools

Installation

{
  "mcpServers": {
    "aip-proxy": {
      "command": "go",
      "args": ["run", "."],
      "env": {
        "AIP_REGISTRY_URL": "https://registry.aip.io"
      }
    }
  }
}

FAQ

How does AIP differ from standard MCP security?

While MCP provides tool access, AIP adds a security layer with agent identity verification, cryptographic authentication tokens, policy-based authorization, and comprehensive audit logging that standard MCP lacks.

What prevents an agent from bypassing the AIP proxy?

AIP operates as a transparent proxy between the AI client and MCP server. All tool calls must pass through it, and the proxy verifies the cryptographic signature of each Agent Authentication Token before allowing execution.

Compare agentidentityprotocol with