Helmdesk — Helpdesk, Email & Knowledge Base for Developers

The FireScraper MCP server lets AI agents — Claude, Cursor, Windsurf, and any Model Context Protocol client — crawl websites and pull clean, LLM-ready text directly from your conversation. No glue code and no separate scripts: your agent gets crawling tools built in.

Before you start

You'll need a FireScraper API key. Create one at firescraper.com under Settings → API Keys. Keys start with fsk_. New accounts include 1,000 free crawl units — no credit card required.

The server runs over stdio via npx, so there is nothing to install globally. You only need Node.js 18 or newer.

Add the server to your client

Claude Code

claude mcp add firescraper --env FIRESCRAPER_API_KEY=fsk_your_key -- npx -y @firescraper/mcp

Claude Desktop, Cursor, or Windsurf

Add this to your MCP configuration file (for example claude_desktop_config.json or ~/.cursor/mcp.json):

json

{
  "mcpServers": {
    "firescraper": {
      "command": "npx",
      "args": ["-y", "@firescraper/mcp"],
      "env": {
        "FIRESCRAPER_API_KEY": "fsk_your_key"
      }
    }
  }
}

Restart the client. You should now see the FireScraper tools available in the conversation.

Available tools

| Tool | What it does |

| --- | --- |

| firescraper_scrape_and_wait | Crawl one or more URLs and return the extracted text in a single call. The easiest place to start. |

| firescraper_scrape | Start a crawl and return a session ID immediately — for large jobs, webhooks, and structured extraction. |

| firescraper_get_session | Check a crawl's status and page counts. |

| firescraper_list_results | List the export files available for a session. |

| firescraper_get_results | Download a session's output in a chosen format (markdown, json, jsonl, documents, chunks, extracted, and more). |

Example

Once the server is connected, ask your agent in plain language:

> "Crawl https://docs.stripe.com one level deep and summarize the webhooks guide."

Behind the scenes the agent calls firescraper_scrape_and_wait and receives clean Markdown for every page, ready to reason over.

Configuration

| Environment variable | Required | Description |

| --- | --- | --- |

| FIRESCRAPER_API_KEY | Yes | Your fsk_ API key. |

| FIRESCRAPER_BASE_URL | No | Override the API base URL (defaults to https://firescraper.com). |

Troubleshooting

"FIRESCRAPER_API_KEY is not set" — make sure the key is provided in the env block of your MCP config (or via --env for Claude Code) and that it starts with fsk_.

Tools don't appear — fully restart your MCP client after editing the config, and confirm Node.js 18+ is installed with node --version.

Large crawls time out — use firescraper_scrape to start the job, then poll with firescraper_get_session and fetch output with firescraper_get_results instead of firescraper_scrape_and_wait.

Using the REST API

Python SDK: Installation, Usage, and LangChain Integration

Using the FireScraper MCP server with Claude, Cursor, and Windsurf