🔥 omp-firecrawl — Firecrawl extension for Oh My Pi

@neuralmux/omp-firecrawl is a Firecrawl extension for Oh My Pi. It gives your agent native tools for scraping, crawling, mapping, searching, batch scraping, and structured extraction — backed by the @mendable/firecrawl-js SDK.

✨ Features

• 🪄 Drop-in replacement for built-in fetch/browse tools. Returns clean LLM-ready markdown by default.
• 🌐 Web search (web, news, images) with optional inline scraping.
• 🗺️ Site mapping to discover URLs before crawling.
• 🕷️ Async crawl jobs with status, cancel, and pagination.
• 📚 Batch scrape across many URLs in a single job.
• 🧩 Structured extraction via JSON Schema.
• 🔒 Never logs or displays your API key.
• ⚡ Powered by the SDK — supports JS rendering, anti-bot bypass, PDFs, screenshots, actions, and proxy modes.

📦 Install

omp install npm:@neuralmux/omp-firecrawl

Try without installing permanently:

FIRECRAWL_API_KEY=fc-... omp -e npm:@neuralmux/omp-firecrawl

Try locally from a checkout:

FIRECRAWL_API_KEY=fc-... omp -e ./

⚙️ Configuration

Set a Firecrawl API key before starting omp (get one at firecrawl.dev):

export FIRECRAWL_API_KEY=fc-your-key

Optional overrides:

Variable	Default	Purpose
FIRECRAWL_API_URL	https://api.firecrawl.dev	Custom API endpoint (self-hosted / proxy). FIRECRAWL_BASE_URL also accepted.
FIRECRAWL_TIMEOUT_MS	(SDK default)	Per-request timeout in milliseconds.
FIRECRAWL_MAX_RETRIES	(SDK default)	Automatic retry budget for transient failures.

The extension never logs or displays the API key.

🛠️ Tools

Tool	What it does
firecrawl_scrape	Scrape one URL into markdown, HTML, links, screenshots, or structured JSON.
firecrawl_search	Search the web (web, news, images) and optionally scrape each result.
firecrawl_map	Discover URLs for a site quickly (sitemap-aware).
firecrawl_crawl	Start an async site crawl job.
firecrawl_crawl_status	Poll a crawl job, with pagination.
firecrawl_crawl_cancel	Cancel a running crawl job.
firecrawl_batch_scrape	Start an async job scraping many URLs at once.
firecrawl_batch_scrape_status	Poll a batch scrape job.
firecrawl_extract	Multi-URL structured extraction with a shared schema or prompt.

All tools fail with a clear configuration error when FIRECRAWL_API_KEY is missing.

💬 Command

/firecrawl

Reports whether the extension sees an API key and which API URL it will call.

🚀 Examples

Scrape a page as markdown:

{
  "url": "https://www.firecrawl.dev/blog",
  "formats": ["markdown"]
}

Search the web with inline scraping:

{
  "query": "Oh My Pi extensions",
  "limit": 5,
  "scrapeOptions": { "formats": ["markdown"] }
}

Map a site:

{
  "url": "https://docs.firecrawl.dev",
  "limit": 50
}

Start a crawl with markdown extraction:

{
  "url": "https://docs.firecrawl.dev",
  "limit": 25,
  "scrapeOptions": { "formats": ["markdown"] }
}

Structured extraction from a single page:

{
  "url": "https://news.ycombinator.com",
  "formats": [
    {
      "type": "json",
      "prompt": "Extract the top 5 stories",
      "schema": {
        "type": "object",
        "properties": {
          "stories": {
            "type": "array",
            "items": {
              "type": "object",
              "properties": {
                "title": { "type": "string" },
                "url": { "type": "string" },
                "points": { "type": "number" }
              }
            }
          }
        }
      }
    }
  ]
}

🧠 Use cases

• Pull docs into your context for grounded code generation.
• Audit competitor pricing or feature pages.
• Crawl a site before migration or content review.
• Search + scrape in a single agent loop for research tasks.
• Structured extraction from product catalogs, listings, or filings.

🗂️ Package layout

omp-firecrawl/
├── src/
│   ├── index.ts              ← entry point + /firecrawl command
│   ├── helpers.ts            ← shared: client init, status, abort, result formatting
│   └── tools/
│       ├── scrape.ts
│       ├── search.ts
│       ├── map.ts
│       ├── crawl.ts          ← crawl + status + cancel
│       ├── batch-scrape.ts   ← batch + status
│       └── extract.ts
├── package.json
├── tsconfig.json
├── biome.json
├── README.md
└── LICENSE

The extension ships as TypeScript source. omp loads it via Bun's native import() — no build step required.

🔗 Related

• Firecrawl — the web data API powering this extension.
• @mendable/firecrawl-js — official JavaScript / TypeScript SDK.
• Oh My Pi — the agent runtime that loads this extension.
• Firecrawl on GitHub — open-source engine.

📄 License

MIT. See LICENSE.