Firecrawl CLI

Use the installed CLI for all supported agent-side Firecrawl work. This skill starts after onboarding: firecrawl must already be installed and authenticated.

Preconditions and authentication

Before an action, run command -v firecrawl, firecrawl --version, and firecrawl --status, then inspect firecrawl --help and the selected command's installed help. Installed help is the syntax source of truth.

If the executable is absent or authentication is invalid, stop or enter Path D. Never install, upgrade, invoke npx, create an account, reproduce raw PKCE, or change transport. Never place API keys in argv (including --api-key), command strings, shell history, logs, evidence, or source control. Use the CLI's secure credential store or an inherited environment variable supplied by the approved secret mechanism; never echo it.

Choose one intent path

Path	Select when	Required behavior
Path A — current-session web data	The current task needs live web data.	Use the narrowest CLI command. A known URL starts with scrape. Do not add an SDK.
Path B — shipped application integration	Firecrawl will run in application, service, script, agent, or pipeline code after this session.	Load references/app-integration.md only for this path. An official Firecrawl SDK may exist only in shipped code; agent research, validation, and smoke requests remain CLI-based.
Path C — repeatable deliverable	The result is a brief, report, dataset, or other artifact powered by web evidence.	Collect through the CLI, retain scrubbed source URLs or job identifiers, and synthesize the artifact.
Path D — exceptional recovery	The installed/authenticated precondition failed, or an installed CLI job failed and remains unresolved after ordinary command help.	Diagnose and remediate only through the CLI, then retry the smallest failed operation.