Turn any website into clean, LLM-ready markdown or structured data — directly from Grok Build.
This plugin bundles Firecrawl's hosted MCP server plus a set of skills, giving Grok Build the ability to scrape, search, crawl, map, extract, and monitor the web. Installing the plugin wires up the MCP automatically — the first web action signs you in through your browser, with no API key to paste.
The hosted MCP is the primary way this plugin works — its native firecrawl_* tools are the default execution path for every web operation. The Firecrawl CLI is an optional fallback, used only when the MCP is unavailable or for local-file workflows (download/parse) that a remote server can't perform.
- Search - Web search with optional scraping of results (supports web, news, and image sources)
- Scrape - Extract clean markdown content from any webpage, with JavaScript rendering
- Map - Discover all URLs on a website
- Crawl - Extract content from entire websites
- Interact - Drive a live browser session on a scraped page — click, fill forms, navigate flows, handle pagination/infinite scroll, and log in
All operations include automatic JavaScript rendering, anti-bot handling, and proxy rotation.
In Grok Build, run /plugin and search for firecrawl, then select it to install. This installs both the skills and the bundled Firecrawl MCP server.
The first time Grok uses a Firecrawl tool, you'll be prompted to authorize in your browser (OAuth). Approve it once — no API key to copy or paste. You can also check the connection any time with /mcp.
That's it. No global npm install, no manual key management.
The bundled MCP covers the full remote toolset (scrape, search, crawl, map, extract, agent, interact, monitor). The one capability it can't provide is parsing files on your local disk — for that, install the Firecrawl CLI:
npm install -g firecrawl-cli
firecrawl login --browser # or: firecrawl login --api-key "fc-YOUR-API-KEY"The firecrawl-cli and firecrawl-parse skills use the CLI when present. Get your free API key at: https://firecrawl.dev/app/api-keys
Once installed, Grok Build automatically uses Firecrawl's MCP tools for web tasks — no extra setup. Just ask naturally:
Search the web:
Search for "best practices for React testing" and compile the key recommendations
Scrape a page:
Scrape https://docs.firecrawl.dev/introduction and summarize the key points
Discover site structure:
Map all URLs on https://firecrawl.dev
Research a topic:
Research the latest developments in AI agents and give me a summary
By default the plugin uses the bundled hosted MCP. The Firecrawl CLI is only needed when the MCP is unavailable or for local-file workflows (download/parse). When used, it runs these commands:
| Command | Description |
|---|---|
firecrawl search "query" |
Search the web (supports --sources, --scrape, --tbs for time filters) |
firecrawl scrape <url> |
Scrape a single page to markdown |
firecrawl map <url> |
Discover all URLs on a site |
firecrawl interact |
Interact with a scraped page — click, fill forms, navigate (operates on a scrape ID) |
firecrawl --status |
Check auth status, concurrency, and credits |
Results are saved to a .firecrawl/ directory in your project to keep Grok Build's context window clean:
.firecrawl/search-react_server_components.json
.firecrawl/docs.github.com-actions-overview.md
.firecrawl/firecrawl.dev.md
The default MCP path needs no configuration — it authenticates via browser OAuth on first use. These variables apply only to the optional CLI / self-hosted paths:
| Variable | Required | Description |
|---|---|---|
FIRECRAWL_API_KEY |
CLI-only | Needed only for CLI auth paths (firecrawl login --api-key or env-var auth); not required for the bundled MCP |
FIRECRAWL_API_URL |
No | Custom API endpoint (for self-hosted instances) |
Firecrawl can be self-hosted. Set FIRECRAWL_API_URL to point to your instance:
export FIRECRAWL_API_URL=https://your-firecrawl-instance.comSee the Firecrawl documentation for self-hosting instructions.
This plugin is licensed under AGPL-3.0, consistent with Firecrawl's open-source license.