> ## Documentation Index
> Fetch the complete documentation index at: https://docs.2027.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# MCP Server

> Connect 2027.dev to Claude Code, Cursor, or any MCP client

# MCP Server

The 2027.dev MCP server lets you query eval reports, request new evals, and track results directly from your AI coding tools.

<Info>
  **No npm install needed.** The MCP server uses HTTP Streamable transport — just add the URL and your API key to your MCP client config.
</Info>

## Setup

### Get your API key

1. Go to [2027.dev/evals](https://2027.dev/evals) and sign in
2. Navigate to **Settings** → **API Keys**
3. Create a new key and copy it (starts with `evals_`)

### Claude Code (one command)

Run this in your terminal — replace `evals_YOUR_API_KEY` with your actual key:

```bash theme={null}
claude mcp add 2027-evals --transport http https://2027.dev/evals/api/mcp --header "Authorization: Bearer evals_YOUR_API_KEY"
```

That's it. Run `claude mcp list` to verify the server is connected.

<Accordion title="Manual config file setup">
  Alternatively, add to your Claude Code MCP config (`~/.claude/claude_desktop_config.json`):

  ```json theme={null}
  {
    "mcpServers": {
      "2027-evals": {
        "url": "https://2027.dev/evals/api/mcp",
        "headers": {
          "Authorization": "Bearer evals_YOUR_API_KEY"
        }
      }
    }
  }
  ```
</Accordion>

### Cursor

Add to your Cursor MCP settings (`.cursor/mcp.json`):

```json theme={null}
{
  "mcpServers": {
    "2027-evals": {
      "url": "https://2027.dev/evals/api/mcp",
      "headers": {
        "Authorization": "Bearer evals_YOUR_API_KEY"
      }
    }
  }
}
```

### Any MCP client

The server uses HTTP Streamable transport at:

```
POST https://2027.dev/evals/api/mcp
Authorization: Bearer evals_YOUR_API_KEY
```

## Available tools

| Tool                  | Description                                               |
| --------------------- | --------------------------------------------------------- |
| `list_reports`        | List eval reports with filtering and pagination           |
| `get_report`          | Full report details — scores, dimensions, recommendations |
| `get_session_trace`   | Agent conversation replay with tool calls                 |
| `get_friction_points` | Extract friction points from a report                     |
| `compare_reports`     | Side-by-side comparison of multiple reports               |
| `get_trends`          | Score history over time                                   |
| `list_prompts`        | Available eval prompts for your org                       |
| `request_eval`        | Trigger a new evaluation run                              |
| `list_eval_requests`  | Track eval request status                                 |

## Example usage

Once connected, you can ask your AI assistant things like:

* "Show me my latest eval reports"
* "What friction points did the last eval find?"
* "Compare the last two reports for my product"
* "Request a new eval for my docs"
* "Show me the session trace for report xyz"