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

# MCP server

> Give AI coding tools the ability to read the web with the Reader MCP server.

The Reader MCP server lets AI coding tools scrape, crawl, extract, and discover web content directly. Install it once and your AI assistant can read any URL, crawl entire sites, or extract structured data on demand.

Published as `@vakra-dev/reader-mcp` on npm. Works with any tool that supports the [Model Context Protocol](https://modelcontextprotocol.io).

## Setup

You need a Reader API key. Get one at [console.reader.dev](https://console.reader.dev).

### Claude Code

```bash theme={null}
claude mcp add reader -- npx -y @vakra-dev/reader-mcp
```

Then set your API key:

```bash theme={null}
export READER_API_KEY=rdr_your_key_here
```

### Claude Desktop

Add to your `claude_desktop_config.json`:

```json theme={null}
{
  "mcpServers": {
    "reader": {
      "command": "npx",
      "args": ["-y", "@vakra-dev/reader-mcp"],
      "env": {
        "READER_API_KEY": "rdr_your_key_here"
      }
    }
  }
}
```

### Cursor

Add to your Cursor MCP settings (Settings > Features > MCP Servers):

```json theme={null}
{
  "mcpServers": {
    "reader": {
      "command": "npx",
      "args": ["-y", "@vakra-dev/reader-mcp"],
      "env": {
        "READER_API_KEY": "rdr_your_key_here"
      }
    }
  }
}
```

### VS Code

Add to your `.vscode/mcp.json` or user settings:

```json theme={null}
{
  "mcpServers": {
    "reader": {
      "command": "npx",
      "args": ["-y", "@vakra-dev/reader-mcp"],
      "env": {
        "READER_API_KEY": "rdr_your_key_here"
      }
    }
  }
}
```

### Windsurf

Add to your `~/.codeium/windsurf/mcp_config.json`:

```json theme={null}
{
  "mcpServers": {
    "reader": {
      "command": "npx",
      "args": ["-y", "@vakra-dev/reader-mcp"],
      "env": {
        "READER_API_KEY": "rdr_your_key_here"
      }
    }
  }
}
```

### Codex

Add to your `~/.codex/config.toml`:

```toml theme={null}
[mcp_servers.reader]
command = "npx"
args = ["-y", "@vakra-dev/reader-mcp"]

[mcp_servers.reader.env]
READER_API_KEY = "rdr_your_key_here"
```

### OpenCode

Add to your `opencode.jsonc`:

```jsonc theme={null}
{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "reader": {
      "type": "local",
      "command": ["npx", "-y", "@vakra-dev/reader-mcp"],
      "enabled": true,
      "environment": {
        "READER_API_KEY": "rdr_your_key_here"
      }
    }
  }
}
```

## Available tools

Once connected, your AI assistant has access to these tools:

| Tool                    | Description                                                                           |
| ----------------------- | ------------------------------------------------------------------------------------- |
| `reader_scrape`         | Scrape a URL and return its content as markdown, HTML, or screenshot                  |
| `reader_batch_scrape`   | Scrape multiple URLs in parallel                                                      |
| `reader_crawl`          | Crawl a website, discovering and scraping all linked pages                            |
| `reader_discover`       | Discover all URLs on a website without scraping content (fast, 1 credit flat)         |
| `reader_extract`        | Scrape a URL and extract structured data using AI (schema or natural language prompt) |
| `reader_job_status`     | Check the status and results of an async job                                          |
| `reader_cancel_job`     | Cancel a running async job                                                            |
| `reader_session_create` | Create a browser session with a CDP WebSocket endpoint                                |
| `reader_credits`        | Check your credit balance and usage                                                   |

## Example prompts

Once the MCP server is connected, you can ask your AI assistant things like:

* "Scrape [https://example.com/docs](https://example.com/docs) and summarize the key points"
* "Crawl [https://docs.example.com](https://docs.example.com) up to depth 3 and find all pages about authentication"
* "Discover all URLs on [https://example.com](https://example.com) and list the ones about pricing"
* "Extract the product name, price, and features from [https://example.com/product](https://example.com/product)"
* "Check my Reader credit balance"

The AI assistant will automatically call the appropriate Reader tool and return the results.

## Self-hosted

If you're running a self-hosted Reader instance, set the `READER_API_URL` environment variable alongside your API key:

```json theme={null}
{
  "mcpServers": {
    "reader": {
      "command": "npx",
      "args": ["-y", "@vakra-dev/reader-mcp"],
      "env": {
        "READER_API_KEY": "rdr_your_key_here",
        "READER_API_URL": "https://your-reader-instance.com"
      }
    }
  }
}
```

## Pricing

MCP tools use the same credit system as the API:

| Operation         | Credits                        |
| ----------------- | ------------------------------ |
| Scrape (standard) | 1 per page                     |
| Scrape (premium)  | 3 per page                     |
| Extract           | +2 per page (on top of scrape) |
| Crawl             | 1 per page                     |
| Discover          | 1 per job (flat)               |
| Browser session   | 1 per minute                   |

## Next

* [Reader as an agent tool](/home/guides/llm/agent-tool): Build your own tool calling integration
* [Structured extraction](/home/guides/llm/structured-extraction): Extract structured data from pages
* [RAG pipeline](/home/guides/llm/rag): Build retrieval-augmented generation with Reader