home / mcp / agntor mcp server

Agntor MCP Server

Provides an MCP server for agent discovery, certification, trust scoring, and governance with hosted and local development options.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "agntor-mcp": {
      "url": "https://mcp.agntor.com/mcp",
      "headers": {
        "PORT": "3100",
        "NODE_ENV": "production",
        "AGNTOR_API_KEY": "YOUR_API_KEY",
        "AGNTOR_SECRET_KEY": "your-secret"
      }
    }
  }
}

You can run and consume the Agntor MCP Server to enable secure agent discovery, certification, and trusted interactions within your applications. This server provides tools to verify agent status, compute trust scores, issue signed tickets, manage registrations, and enforce safety checks, all accessible via HTTP or local stdio configurations for development and production workflows.

How to use

You will connect MCP clients to the Agntor MCP Server either over HTTP for hosted endpoints or via local stdio for development and testing. From there you can query agent certification, check trust levels, issue audit tickets, search agents, and perform emergency actions like kill switches. Use the hosted MCP URL for production deployments and run the local server in development to test integrations before going live.

How to install

Prerequisites you need before installation: a modern Node.js runtime (recommended LTS) and npm. Ensure you have network access to install global packages.

# Install the MCP server globally
npm install -g @agntor/mcp

Additional configuration and usage notes

You can access a hosted MCP endpoint or run a local server for development. The hosted endpoint is available at the MCP URL below and can be secured with an API key if enabled.

Hosted MCP URL (production): https://mcp.agntor.com/mcp. If enabled, include the API key header X-AGNTOR-API-KEY: <your_key> when calling endpoints.

Environment variables and runtime options

Use the following environment variables to configure the server in production and development environments.

- PORT: HTTP server port (default 3100) - AGNTOR_SECRET_KEY: JWT signing key (required for secure operations) - AGNTOR_API_KEY: Optional API key for hosted MCP access - NODE_ENV: Environment indicator (development, production)

Tools and endpoints you can use

The server exposes a suite of tools to manage and assess agents. You can verify certification status, compute trust scores, generate signed audit tickets, search for agents, activate kill switches, and implement safety guards for inputs and tool usage.

Troubleshooting and notes

If you encounter connectivity issues, verify that the MCP HTTP URL is reachable and that any required API keys are correctly configured. For local development, ensure you are running the server from the correct package context and that environment variables are provided for the production-like setup if needed.

Security considerations

Rotate the AGNTOR_SECRET_KEY regularly and enable TLS in production. Protect the MCP endpoints with proper authentication and implement rate limiting to prevent abuse. Monitor tool usage and latency to detect anomalies.

Next steps

1. Set up a custom agent registry backend (database-backed) for persistence. 2. Integrate with downstream systems (x402 payment gateways, external identity providers). 3. Add webhooks for kill switch notifications and auditing. 4. Develop a behavioral scoring model to enhance trust assessments. 5. Build an onboarding dashboard for agents to manage their certifications.

Examples and quick starts

Run the server in development mode to iterate quickly. Use your environment file or shell exports to provide the necessary keys and settings, then start the server and connect MCP clients to test tool calls.

Available tools

is_agent_certified

Return a boolean indicating whether a given agent is certified, along with audit level, expiry, and kill switch status.

get_trust_score

Compute a comprehensive trust score for an agent, including multiple contributing factors and a recommended action.

issue_audit_ticket

Generate a signed JWT ticket for a specific agent with a defined validity period for x402 transactions.

query_agents

Search for agents by criteria such as trust score, audit level, and capabilities, returning matched results.

activate_kill_switch

Emergency action to disable an agent, with optional ticket revocation and timestamp.

guard_input

Validate inputs to detect prompt injections or unsafe instructions and classify potential violations.

redact_output

Redact sensitive information from outputs according to defined redaction policies.

guard_tool

Authorize or block the execution of tools based on policy and risk assessments.

get_agent_registration

Provide an ERC-8004 compatible registration file for agent discovery