Firecrawl

A Firecrawl integration gives assistants tools to scrape URLs, search the web, and parse files into structured content. A deployment may fund those calls with its platform credential, or your organization can connect its own key.

Use Firecrawl when an assistant needs fresh web content or document extraction before an assistant step reasons over the result.

Capability map

Connect Firecrawl

If the catalog marks Firecrawl ready through a platform credential, the action can run without an organization-local key. Connect your own key when usage should be billed to your Firecrawl account.

A platform credential is not a connected account, so there is no account manager or assistant grant to configure for it. Tool selection and roles still apply.

For your own key, use Account manager to choose You or Your team, then grant the saved connection under Assistant access. When several Firecrawl keys are eligible, set the one that should supply quota in Account preferences or on the routine. Capability keys are pinned by configuration; the assistant does not choose a billing key on each scrape.

Actions

firecrawl.parse
firecrawl.scrape
firecrawl.search

Use firecrawl.scrape for a known URL, firecrawl.search for discovery, and firecrawl.parse when the input is a file or document. Save large outputs as artifacts when later steps need to inspect them.

By default firecrawl.search returns only result metadata (title, URL, and description). It does not fetch page content. Let the assistant decide which results are worth reading and call firecrawl.scrape on those URLs. To pull page content inline with the search instead, pass scrapeOptions (for example { "formats": [{ "type": "markdown" }] }) on the search action.

Events

Firecrawl does not register source events.

Next

• Review connected-key access in Connect accounts.
• Save extracted content as artifacts.
• Grant Firecrawl actions to an assistant from actions.