@faultr/mcp-server

A Model Context Protocol (MCP) server for Faultr, providing direct integration with the Faultr agentic stress-testing and compliance API. This server exposes the full Faultr API surface as tools, enabling AI IDEs and agents (Cursor, Claude Desktop, Windsurf, Claude Code, JetBrains) to seamlessly discover, run, and report on evaluations.

Setup & Authentication

This MCP server communicates directly with the Faultr API (defaults to https://app.faultr.ai). You must provide your CLI API Key via an environment variable:

• FAULTR_API_KEY: Required. Your Faultr API Key.
• FAULTR_BASE_URL: Optional. Default https://app.faultr.ai.

IDE Integrations

You can use this package via npx locally. Update your respective IDE configurations with the following snippets:

Cursor

Add to .cursor/mcp.json in your project root, or via Cursor Settings > Features > MCP.

{
  "mcpServers": {
    "faultr": {
      "command": "npx",
      "args": ["-y", "@faultr/mcp-server"],
      "env": {
        "FAULTR_API_KEY": "ft_pro_..."
      }
    }
  }
}

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "faultr": {
      "command": "npx",
      "args": ["-y", "@faultr/mcp-server"],
      "env": {
        "FAULTR_API_KEY": "ft_pro_..."
      }
    }
  }
}

Windsurf / Claude Code / Generic Stdio

Use the identical format from above for any client supporting MCP via STDIO.

Available Tools

1. faultr_list_scenarios: Discover and filter the available compliance test scenarios by domain, difficulty, or protocol.
2. faultr_get_scenario: Get comprehensive details for a specific scenario (trap conditions, criteria, mandates).
3. faultr_evaluate: Run a complex agent trace against scenarios. Supports multi-step action traces and parses the deep JSON findings into readable text for your LLM.
4. faultr_batch_evaluate: Convenience tool to evaluate many scenarios in parallel.
5. faultr_generate_mandates: Automatically create protocol-compliant adversarial fixtures (e.g., AP2 Checkout state constraints).
6. faultr_create_custom_scenario: Build a domain-specific scenario directly into your private library.
7. faultr_generate_custom_scenario: Auto-generate a structured scenario payload from a plain English prompt.
8. faultr_get_report: Retrieve the URL for a highly-detailed HTML compliance report.

---

MCP GitHub Listing Metadata

{
  "name": "faultr-mcp-server",
  "description": "Integration with the Faultr compliance API for agentic stress-testing",
  "vendor": "Faultr AI",
  "language": "TypeScript",
  "transport": ["stdio"],
  "tags": ["testing", "ai", "compliance", "security", "faultr"]
}