Skip to content
MCP server

AgentTwin

Tests AI agent workflows against digital twins with injected faults, run locally via MCP

Runs locallyStatus and history
Install
Pricing
Contact sales
Platforms
CLI, API, self-hosted
Runs
Locally
Version
0.3.27
Listing updated
today

Similar MCP servers

About AgentTwin

AgentTwin is an MCP server for developer tools. It runs locally on your machine and is started by your AI client.

The README describes AgentTwin as Node.js workflow testing for prepared applications and AI agents. Workflows run against synthetic service state, supported failures are injected, and operations, state changes and assertions are recorded for review. Developers can start from the npx agenttwin CLI, which uploads a project map and offers guided tests that run in local Docker, or use the local stdio MCP server to let a coding agent drive world, scenario, execution and inspection tools.

The product is positioned around catching unsafe retries and failure handling before deployment. Its documentation stresses that static discovery is not a runnable test and that no production replica or provider parity is promised. Local execution can access the host filesystem and network, so the site advises reviewing code, fixtures and invocation first.

The offering suits engineering teams that build agents and want repeatable failure tests. Access to the API and the CLI's cloud runs is tied to an account, and the vendor's pricing page lists a single enterprise offering handled through sales.

Install

Run in a terminal

claude mcp add agenttwin -- npx -y agenttwin

Add to claude_desktop_config.json

{
  "mcpServers": {
    "agenttwin": {
      "command": "npx",
      "args": [
        "-y",
        "agenttwin"
      ]
    }
  }
}
Install in CursorOpens Cursor, which asks you to confirm. Or add it by hand:

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "agenttwin": {
      "command": "npx",
      "args": [
        "-y",
        "agenttwin"
      ]
    }
  }
}
Install in VS CodeOpens VS Code, which asks you to confirm. Or add it by hand:

Add to .vscode/mcp.json

{
  "servers": {
    "agenttwin": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "agenttwin"
      ]
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "agenttwin": {
      "command": "npx",
      "args": [
        "-y",
        "agenttwin"
      ]
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "agenttwin": {
      "command": "npx",
      "args": [
        "-y",
        "agenttwin"
      ]
    }
  }
}

Using several servers? Build one config file with the MCP config generator.

What it does

Synthetic service twins
Runs agent workflows against synthetic service state instead of live production systems.
Injected fault testing
Injects supported failures so developers can see how an agent handles them.
Local MCP server
Exposes world, scenario, execution and inspection tools to coding agents over stdio.
Recorded run review
Records operations, state changes and assertions for review and replay.
Node CLI workflow
Uses npx agenttwin to discover project context and prepare tests in a repository.

Good for

  • Testing how an AI agent retries a failing service call before shipping it
  • Replaying a recorded failure to check whether a repair holds
  • Preparing synthetic test workflows from a coding agent inside a repository
  • Writing regression checks for agent workflows that touch external services

Pricing

Contact sales

Single enterprise offering; contact sales for pricing.

Prices change. Check agenttwin.si before you buy.

AgentTwin FAQ

Is there a free tier?

The pricing page lists one enterprise offering, handled through contact with sales. No free plan is described there, though the CLI can run a separate sample test and a credential-free synthetic sandbox is mentioned.

Does the local MCP server run in the cloud?

No. The developer page says it is a local stdio server run with npx agenttwin mcp, not a hosted Streamable HTTP endpoint.

Is the source code available?

The developer page says implementation source is currently private and repository access is not included with account sign-up. Public integration files are in a separate agenttwin-agent-tools repository under MIT licence.

Does the public documentation MCP need credentials?

No. The docs MCP at agenttwin.si/mcp/docs provides search_docs and read_doc without credentials, but it cannot access accounts or run workflows.

How much does AgentTwin cost?

AgentTwin does not publish prices. Contact the company for a quote.

How do I add AgentTwin to Claude Code?

Run this command in your terminal: claude mcp add agenttwin -- npx -y agenttwin

What are the best alternatives to AgentTwin?

The closest alternatives to AgentTwin are AgentTwin Documentation, fieldtwin, Agent Tested, Tenhaw Kit and n8n-mcp.

Where this comes from

  • Description: written from the project's README, the project's website and its MCP Registry entry.
  • Install steps: from the official MCP Registry entry.

How AIPoweredBot checks servers