# Guardrails MCP Server

MCP (Model Context Protocol) server for [Guardrails](https://useguardrails.com) — the conscience layer for autonomous agents.

Allows any MCP-compatible AI agent (Claude, Cursor, Windsurf, etc.) to natively use Guardrails to check user values, make aligned decisions, and respect faith-based preferences.

## Transport

Streamable HTTP (MCP spec 2025-03-26) at `/mcp` endpoint.

## Tools

| Tool | Description |
|------|-------------|
| `create_profile` | Create a new user values profile |
| `get_profile` | Retrieve a user's values profile |
| `update_profile` | Update profile domains |
| `check` | Rule-based compatibility check |
| `decide` | LLM-powered decision (premium, BYOK) |
| `calendar` | Faith-aware calendar/scheduling |
| `get_taxonomies` | List available value domains |
| `list_configurations` | List custom configs |
| `create_configuration` | Create custom config |
| `update_configuration` | Update custom config |
| `delete_configuration` | Delete custom config |

## Agent Configuration

### Claude Desktop / Cursor / Windsurf

Add to your MCP config:

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

Every tool requires your Guardrails `api_key` as a parameter.

## Environment Variables

| Variable | Default | Description |
|----------|---------|-------------|
| `PORT` | `3001` | Server port |
| `GUARDRAILS_API_URL` | `https://api.useguardrails.com` | Guardrails REST API base URL |

## Development

```bash
npm install
npm run dev
```

## Deployment

Deployed as a separate Railway service in the same project as the API.