plompy-engine.Docs
DocumentationFirecrawl

Firecrawl

Firecrawl can discover pages and extract their content. It is an external service: queries and requested URLs leave your computer. The gateway supports an explicitly selected tool subset.

Limited keyless setup#

sh
docker compose exec -T local \
  node scripts/integrations.mjs add firecrawl --mode local \
  --auth keyless --tools firecrawl_search,firecrawl_scrape

Keyless access is subject to the vendor's current restrictions and availability. It is not a promise of an unlimited free plan.

API-key setup#

Choose API-key mode:

sh
docker compose exec -T local \
  node scripts/integrations.mjs add firecrawl --mode local \
  --auth api-key --tools firecrawl_search,firecrawl_scrape

Provide a dedicated key through the private credential workflow. Do not type the key in the command arguments or save it in your client project configuration. The local gateway does not perform Firecrawl OAuth login.

Apply the configuration#

After adding the integration and any required credential, restart the gateway and reconnect the host client so it discovers the updated tools:

sh
docker compose restart local

Use local-browser instead of local if the browser service is running. A configuration change does not update an already running gateway automatically.

Try it#

text
Use the configured Firecrawl search tool to find up to 3 Polish mobile
app development companies. Read official pages and retain their URLs.
Use web_fetch or the allowed scrape tool for the evidence.

Read-only argument surface#

The gateway limits firecrawl_search to a query and bounded result count. firecrawl_scrape accepts a public URL, optional main-content extraction and only Markdown, HTML or links as output formats. firecrawl_map accepts a URL, optional search text and a bounded result count. Browser actions, arbitrary JavaScript and extra upstream arguments are rejected.

Costs and limits#

Your account, credits, rate limits and the vendor's terms apply. Search and scraping can consume credits. No extra provider calls are made merely by reading local configuration status.

Discovery and a real tool call are different checks. Run an explicit live diagnostic, then one small search you approve. Never treat a stored key as proof of a successful search.

The gateway exposes only configured read tools that pass its research policy. It does not automatically enable new crawl, agent or mutation tools when the vendor adds them.

See Firecrawl MCP documentation for provider requirements.

Find your next step

Search titles and guide content.