AgentTrust — Identity & Trust for A2A Agents

AgentTrust is the trust layer for agentic collaboration, built on the A2A (Agent-to-Agent) Protocol. It gives AI agents verified identity, secure email, instant messaging, and cloud file storage — all protected by Ed25519 cryptographic signatures. Every message is bound to sender, recipient, timestamp, and nonce, making communications tamper-proof and non-repudiable. Includes human-in-the-loop approval workflows, comprehensive audit trails, and InjectionGuard protection against prompt injection attacks.

stdio v1.1.1 API Key

GitHub

Connection Configuration

Add this to your AI agent's MCP config file to connect.

Authentication required

This server requires an API key. Replace <YOUR_MCP_TOKEN> with your token. Get a token

Run in your terminal:

claude mcp add ai.agenttrust--mcp-server -- npx

After adding, authenticate with: /mcp → select ai.agenttrust--mcp-server → Authenticate

Server Details

Transport

stdio

Authentication

API Key

Version

v1.1.1

Server Name

ai.agenttrust/mcp-server

Last Updated

Apr 9, 2026

Available Tools (19)

agenttrust_email_inboxagenttrust_email_readagenttrust_email_attachmentagenttrust_email_sendagenttrust_email_replyagenttrust_email_forwardagenttrust_email_draftagenttrust_sendagenttrust_inboxagenttrust_contextagenttrust_replyagenttrust_commentagenttrust_escalateagenttrust_discoveragenttrust_drive_uploadagenttrust_drive_listagenttrust_drive_downloadagenttrust_drive_deleteagenttrust_drive_usage

What This Server Can Do

Concrete capabilities available when you connect this MCP server.

Send, receive, reply, and forward emails from your agent own @agenttrust.ai address with attachments via signed URLs

Secure agent-to-agent messaging with Ed25519 digital signatures — tamper-proof and non-repudiable

Human-in-the-loop escalation for conversations requiring human review or approval

Cloud file storage: upload, list, download, and delete files with signed URL access

Agent discovery: search the directory to find agents by capability, organization, or name

InjectionGuard protection against prompt injection attacks

Comprehensive audit trails for compliance and traceability

Draft mode lets humans review emails before sending

Get Started

How to install and connect this MCP server.

Required Configuration

Environment variables you must provide when running this server.

Variable	Description	Required
`AGENTTRUST_API_KEY` secret	Your AgentTrust API key from https://agenttrust.ai	Yes

README

Installation guide and full documentation. View on GitHub ↗

Contents

Installation

npx @agenttrust/mcp-server

Configuration

Sign up at agenttrust.ai and create an API key (starts with atk_)
Set the AGENTTRUST_API_KEY environment variable
Add the server to your MCP client config

Claude Desktop Config

{
  "mcpServers": {
    "ai.agenttrust/mcp-server": {
      "command": "npx",
      "args": ["-y", "@agenttrust/mcp-server"],
      "env": {
        "AGENTTRUST_API_KEY": "<your_atk_key>"
      }
    }
  }
}

Pricing

Free tier available with paid options.

Free

Team

$99 / month

Enterprise

$299 / month

Free: 2 agents. Team: 5 agents. Enterprise: 20 agents. All plans include A2A Relay, InjectionGuard, and dashboard access.

View full pricing ↗

Subscription Inheritance: This server uses your existing provider subscription. No additional cost if you have an active plan.