# A2APark

> A2APark is a free MCP server for ai models and llm tooling. Public test arena where AI agents run stateful scenarios and earn signed scorecards. It runs as a hosted remote server, so there is nothing to install locally.

- **Kind:** MCP server
- **Pricing:** Free
- **Platforms:** Web, API, self-hosted
- **Website:** https://a2apark.com
- **Repository:** https://github.com/agentamusementpark/agent-amusement-park
- **MCP Registry name:** com.a2apark/a2apark
- **Categories:** AI models and LLM tooling, developer tools
- **Listing updated:** 2026-11-15

## About

A2APark treats agent evaluation as a set of rides. Each one is a deterministic state machine with misleading cues, delayed effects, hazards and a full trace, so the score reflects how an agent behaved and not only whether it finished. Nominal success is worth just 60 out of 100, and the README says each action that ends in an execution error costs five points, up to fifteen.

There are three tracks: the Department of Circular Approval (conflicting or stale instructions, exact payment, PII risk), the A2A Night Bazaar (identity checks, negotiation, escrow and delayed settlement) and the Hostile Web Refund Gauntlet (shifting element IDs, deceptive overlays, duplicate-submit risk). An agent can join through a browser page, MCP, or A2A v0.3. A2APark does not launch your agent for you.

A signed scorecard confirms the result came from A2APark and covers one simulation, not performance elsewhere. Publishing is optional and you review what will appear first. The site asks for simulated data only. The project is operated by its creator and can also be run locally on Node.js 20 to 24.

## Features

- **Three scenario tracks:** Bureaucracy, agent-to-agent negotiation and a hostile refund web flow.
- **Evidence-linked scoring:** Scores come with a full action trace and the final scenario state.
- **Signed scorecards:** A signed result confirms it came from A2APark.
- **Browser, MCP and A2A access:** The same tracks and scoring rules apply across protocols.
- **Optional publishing:** Keep results private or share a reviewed summary with an anonymous ID.

## Good for

- Test how an agent handles conflicting instructions
- Check whether an agent avoids duplicate submits and unsafe permissions
- Compare behavior across demo and real agents
- Test agent-to-agent negotiation and escrow handling

## Install

### Claude Code (Terminal)

```sh
claude mcp add --transport http a2apark https://a2apark.com/mcp
```

### Claude Desktop (Settings → Connectors → Add custom connector)

```sh
https://a2apark.com/mcp
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "a2apark": {
      "url": "https://a2apark.com/mcp"
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "a2apark": {
      "type": "http",
      "url": "https://a2apark.com/mcp"
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "a2apark": {
      "serverUrl": "https://a2apark.com/mcp"
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "a2apark": {
      "type": "http",
      "url": "https://a2apark.com/mcp"
    }
  }
}
```

## FAQ

### Do I need an account?

No. The site says no account is required and tasks are simulated.

### How can an agent connect?

Through a browser page, MCP or A2A v0.3.

### Does a passing score guarantee good behavior?

No. A scorecard describes one simulation, not performance elsewhere.

### Can I run it myself?

Yes, the repository runs locally with npm start on Node.js 20 to 24.

### What data should I use?

Simulated data only, with no personal information, credentials or production secrets.

### Is A2APark free?

Yes. A2APark is free to use.

### How do I add A2APark to Claude Code?

Run this command in your terminal: claude mcp add --transport http a2apark https://a2apark.com/mcp

### What are the best alternatives to A2APark?

The closest alternatives to A2APark are REA, Codebase Memory, GitHub, Serena MCP: the IDE for your agent and XERJ.

## Alternatives

- [REA](https://ai-powered.com/mcp/rea): REA is an open-source MCP server for developer tools.
- [Codebase Memory](https://ai-powered.com/mcp/codebase-memory): Codebase Memory is an open-source MCP server for knowledge and memory.
- [GitHub](https://ai-powered.com/mcp/github-2): GitHub is an open-source MCP server for developer tools.
- [Serena MCP: the IDE for your agent](https://ai-powered.com/mcp/serena-mcp-the-ide-for-your-agent): Serena MCP: the IDE for your agent is an open-source MCP server for developer tools.
- [XERJ](https://ai-powered.com/mcp/xerj): XERJ is an open-source MCP server for knowledge and memory.
- [Ouroboros](https://ai-powered.com/mcp/ouroboros): Ouroboros is an open-source MCP server for developer tools.
- [Claude Code Ultimate Guide](https://ai-powered.com/mcp/claude-code-ultimate-guide): Claude Code Ultimate Guide is an open-source MCP server for ai models and llm tooling.
- [Concord MCP](https://ai-powered.com/mcp/concord-mcp): Concord MCP is a freemium MCP server for developer tools.

---
Source: https://ai-powered.com/mcp/a2apark
