App Store/Web & Automation/Firecrawl

Firecrawl

Name: Firecrawl
Author: Firecrawl

Firecrawl · io.pilot.firecrawl

Turn any website into clean, LLM-ready data.

Web & AutomationLive on catalogue

Install

pilotctl appstore install io.pilot.firecrawl

v0.1.0

Version

Methods

8.6 MB

Size

shareable

Sandbox

macOS · Linux

Platforms

About Firecrawl

Firecrawl turns any website into clean, LLM-ready data. This app exposes the complete Firecrawl v2 API — all 50 operations — as Pilot app-store methods, generated 1:1 from Firecrawl's published OpenAPI spec.

You do not need a Firecrawl account, an API key, an inbox, or a browser. The app is keyless: Pilot's broker verifies your agent identity, meters your usage, and authenticates upstream on your behalf. Install and call.

What you get

- Scrape — one URL to clean markdown, HTML, links, screenshots, or schema-shaped JSON. Handles JS-heavy pages, PDFs and Office documents, geo/proxy emulation, ad-blocking, and PII redaction. - Search — web, news, and image search with query-relevant highlights, domain and time filters, and optional full scraping of every result in the same call. - Map — enumerate every URL on a domain in about a second. The cheap way to scope a crawl before you pay for one. - Crawl — async whole-site extraction with include/exclude path regexes, depth and page limits, subdomain and external-link control, robots.txt handling, and webhooks. Describe the crawl in plain English and preview the generated options for free. - Batch scrape — many known URLs in one job, with per-URL error reporting. - Extract — LLM-powered structured extraction across many pages against your JSON Schema, optionally with web search and source citations. - Agent — autonomous web research. Give a prompt and a credit ceiling; the agent decides where to go and returns structured data. - Interact — drive a real browser with Playwright or natural language, either bound to a scrape job or as a standalone session with persistent profile storage. - Monitor — recurring checks on pages, whole sites, or web-wide searches, with an AI judge that filters noise against a plain-language goal, and webhook/email/Slack alerts. - Research index — search scientific papers, read their full-text passages, expand to related work, and search GitHub issues, PRs, and READMEs. - Support — firecrawl.ask diagnoses a failing job from your account's logs, and firecrawl.docs_search answers Firecrawl questions with citations. - Usage — credits, tokens, queue depth, and recent activity, so an agent can check cost and capacity before committing to an expensive call.

Start with firecrawl.help — it returns every method with its full parameter set, latency class, and measured roundtrip.

Cost control. firecrawl.crawl takes limit and firecrawl.agent takes maxCredits; both default high. Set them. firecrawl.map and firecrawl.crawl_params_preview are the cheap ways to scope work before you spend. Check firecrawl.credit_usage and firecrawl.balance before a large job.

Not exposed: POST /v2/parse (upload a local file for parsing) takes multipart/form-data, which the JSON-in/JSON-out app-store IPC contract cannot express. Use firecrawl.scrape on a public document URL instead — it parses PDF, DOCX, XLSX and friends natively.

Methods · 51

50 methods work on the free plan. The 1 below the divider need an account upgrade (paid plan, invite, or a verified custom domain).

firecrawl.scrape

Scrape one URL into clean LLM-ready markdown (or html/links/screenshot/json).

firecrawl.scrape_status

Get the status and result of a previously started scrape job by id.

firecrawl.scrape_interact

Run code or an AI prompt in the live browser session bound to a scrape job.

firecrawl.scrape_interact_stop

Stop the interactive browser session attached to a scrape job and free its resources.

firecrawl.search

Search the web (and optionally scrape every result in one call).

firecrawl.search_feedback

Rate a search job's quality to improve future Firecrawl search results.

firecrawl.map

Discover every URL on a domain, fast — the cheap way to scope a crawl.

firecrawl.crawl

Start an async crawl of an entire site. Returns a job id — poll firecrawl.crawl_status.

firecrawl.crawl_status

Poll a crawl job: status, completed/total counts, credits used, and the scraped pages.

firecrawl.crawl_cancel

Cancel a running crawl job.

firecrawl.crawl_errors

List the pages a crawl job failed on, plus URLs blocked by robots.txt.

firecrawl.crawl_active

List every crawl currently running for your account.

firecrawl.crawl_params_preview

Dry-run: see the crawl options a natural-language prompt would generate, without spending credits.

firecrawl.batch_scrape

Scrape many known URLs in one async job. Returns a job id — poll firecrawl.batch_scrape_status.

firecrawl.batch_scrape_status

Poll a batch scrape job: progress, credits used, and the scraped pages.

firecrawl.batch_scrape_cancel

Cancel a running batch scrape job.

firecrawl.batch_scrape_errors

List the URLs a batch scrape job failed on, with per-URL error detail.

firecrawl.extract

Extract structured JSON from one or many pages using an LLM and your schema. Async — poll firecrawl.extract_status.

firecrawl.extract_status

Poll an extract job and retrieve the structured data once it completes.

firecrawl.agent

Autonomous web research: give a prompt, the agent decides what to visit and returns structured data. Async — poll firecrawl.agent_status.

firecrawl.agent_status

Poll an agent job and retrieve its result once it completes.

firecrawl.agent_cancel

Cancel a running agent job and stop it spending credits.

firecrawl.interact_create

Start a standalone browser session you drive with code — no prior scrape needed.

firecrawl.interact_list

List your standalone interact browser sessions.

firecrawl.interact_execute

Run Playwright or agent-browser code inside a standalone interact session.

firecrawl.interact_delete

Destroy a standalone interact session and release its browser.

firecrawl.monitor_create

Create a recurring check that watches pages, a whole site, or search results and alerts on meaningful change.

firecrawl.monitor_list

List your monitors.

firecrawl.monitor_get

Get one monitor's full configuration.

firecrawl.monitor_update

Update a monitor's schedule, targets, goal, or status (active/paused).

firecrawl.monitor_delete

Delete a monitor and stop its recurring checks.

firecrawl.monitor_run

Trigger a monitor check immediately instead of waiting for its schedule.

firecrawl.monitor_checks

List a monitor's past checks and what changed in each.

firecrawl.monitor_check

Get one monitor check in full, with page-level diffs.

firecrawl.research_papers

Search a purpose-built scientific paper index by natural-language query.

firecrawl.research_paper

Inspect a paper's metadata, or pass query to read its top matching full-text passages.

firecrawl.research_related

Expand a seed paper to related work — citers, references, or semantically similar papers.

firecrawl.research_github

Search GitHub issues, pull requests, discussions, and READMEs by natural-language query.

firecrawl.ask

Diagnose a failing Firecrawl call with an AI support agent — pass the jobId and get a fix.

firecrawl.docs_search

Answer a "how does Firecrawl do X?" question from the official docs, with citations.

firecrawl.feedback

Submit quality feedback for a completed search, scrape, parse, or map job.

firecrawl.credit_usage

Remaining credits and the current billing period.

firecrawl.credit_usage_historical

Historical credit usage over time, optionally broken down by API key.

firecrawl.token_usage

Remaining LLM tokens for extract-family operations.

firecrawl.token_usage_historical

Historical token usage over time, optionally broken down by API key.

firecrawl.queue_status

Your scrape queue depth, active jobs, and max concurrency — check before a large batch.

firecrawl.activity

Recent API activity (last 24h) with job ids you can feed to the *_status methods.

firecrawl.threat_protection

Read the account's threat-protection policy (URL risk checks, allow/deny lists).

firecrawl.balance

Your remaining credits for this app, read free from the broker's per-user ledger — returns {"credits_remaining":<n>,"credits_seed":<n>,"scope":"per-pilot-user"}. This is YOUR balance. No partner API call, no charge. Check it before an expensive job.

firecrawl.help

Discovery: every method with params, kind, and latency class.

Requires an upgrade · 1

firecrawl.threat_protection_updateupgrade

Replace the account's threat-protection policy. Full-document update — unspecified fields reset to defaults.

Not available on the Pilot plan. Threat-protection policy applies to a whole Firecrawl team rather than a single caller, so it is not something an individual Pilot user can set. Upstream it is also an enterprise feature restricted to team admins.

Full usage demo

When you need content from the live web: read a page as clean markdown, search the web, enumerate a site's URLs, crawl a whole domain, or extract structured data from pages.

Run this first — Read any page as clean, LLM-ready markdown1 credit

Call

pilotctl appstore call io.pilot.firecrawl firecrawl.scrape '{"url":"https://example.com"}'

→

{"success":true,"data":{"markdown":"# Example Domain\n\nThis domain is for use in documentation examples...","metadata":{"title":"Example Domain","creditsUsed":1}}}

Worked examples

Check your budget first (free read)0 credits (read)

Call

pilotctl appstore call io.pilot.firecrawl firecrawl.balance '{}'

→{"credits_remaining":1000,"credits_seed":1000,"scope":"per-pilot-user"}

Search the web and get real content back2 credits

Call

pilotctl appstore call io.pilot.firecrawl firecrawl.search '{"query":"open source web crawlers","limit":3}'

→{"success":true,"data":{"web":[{"url":"https://...","title":"...","description":"..."}]},"id":"019fc9b8-..."}

Enumerate a site's URLs before paying to crawl it1 credit

Call

pilotctl appstore call io.pilot.firecrawl firecrawl.map '{"url":"https://docs.firecrawl.dev","limit":50}'

→{"success":true,"links":[{"url":"https://docs.firecrawl.dev/...","title":"..."}]}

Always map before a crawl: crawl defaults to 10000 pages.

Crawl a site — async, returns a job id20 credits

Call

pilotctl appstore call io.pilot.firecrawl firecrawl.crawl '{"url":"https://docs.firecrawl.dev","limit":10}'

→{"success":true,"id":"a1b2c3d4-...","url":"https://api.firecrawl.dev/v2/crawl/a1b2c3d4-..."}

SET limit — it defaults to 10000 pages.

Poll the crawl until it finishes (free read)0 credits (read)

Call

pilotctl appstore call io.pilot.firecrawl firecrawl.crawl_status '{"id":"a1b2c3d4-..."}'

→{"status":"completed","total":10,"completed":10,"creditsUsed":10,"data":[{"markdown":"...","metadata":{}}]}

What it costs

1000 credits per Pilot userfree budget · billed in Firecrawl credits

Operation	Price	Notes
`scrape / map / docs_search`	1 credit	per call
`search`	2 credits	per query
`ask`	3 credits	per diagnosis
`interact_create / monitor_create`	5 credits	per session or monitor
`crawl / batch_scrape / extract`	20 credits	charged when the job is accepted; a large crawl costs more as it runs
`agent`	50 credits	cap it with maxCredits
`all _status, _errors, list and usage reads`	0 credits	reads are free

This demo spends 24 of your 1000 credits (quickstart scrape 1 + search 2 + map 1 + crawl 20; both reads are free).

Check balance

pilotctl appstore call io.pilot.firecrawl firecrawl.balance '{}'

Good to know

You get 1000 credits and 2 concurrent calls. Check firecrawl.balance before a big job.
429 means you already have 2 calls in flight — wait, or free a slot. Browser/interact sessions hold one until stopped.
crawl defaults to limit 10000 and agent to maxCredits 2500 — always set them.
crawl/batch_scrape/extract/agent are async: they return a job id, poll the matching *_status.
404 on someone else's job id is intentional — jobs are isolated per Pilot user.
firecrawl.parse is not exposed (multipart upload); scrape a public PDF/DOCX URL instead.

io.pilot.firecrawl firecrawl.help '{}'

What’s New

v0.1.0Latest

—

Released v0.1.0

Platform Compatibility

macOSApple Silicon

5.0 MBSupported

macOSIntel

5.0 MBSupported

Linuxarm64

4.5 MBSupported

Linuxamd64

4.5 MBSupported

More in Web & Automation

Plainweb

Pilot Protocol

Any web page as clean Markdown — no HTML, no JS, one call

macOS · Linux

Otto

Telepat

Drive real Chrome tabs from an agent — extract, automate, screenshot, no headless farm

macOS · Linux

Bowmark

Bowmark AI

Navigation cheatsheets for public websites, so agents run cheaper, faster, and more accurately

macOS · Linux

Firecrawl

About Firecrawl

Methods · 51

Full usage demo

Worked examples

What it costs

Good to know

Next

What’s New

Platform Compatibility

More in Web & Automation