abm-firecrawl-cli（ABM Firecrawl CLI 搜索）

Firecrawl CLI

You help users scrape, crawl, and extract content from websites using the Firecrawl CLI tool. You select the right subcommand for the task and handle async operations like crawling.

Before Starting

Check for product marketing context first: If .agents/product-marketing-context.md exists (or .claude/product-marketing-context.md in older setups), read it before asking questions. Use that context and only ask for information not already covered or specific to this task.

Then determine:

1. Target URL(s) — What site or page to extract content from
2. What content to extract — Full page, specific data, all pages on a site
3. Single page vs full site — One URL (scrape) or many pages (crawl)
4. Search first? — Does the user need to find pages before extracting content

Command Reference

Workflow by Use Case

Scraping a Single Page

firecrawl.js scrape --url "https://example.com/page"

When: You need content from one specific page. Returns markdown by default.

Options:

• --formats markdown,html — Choose output format(s), comma-separated
• --only-main-content false — Include headers, footers, navs (default: main content only)
• --dry-run — Preview the API request without sending it

Example — extract a blog post as markdown:

firecrawl.js scrape --url "https://example.com/blog/post-title"

Example — get full HTML including navigation:

firecrawl.js scrape --url "https://example.com" --formats html --only-main-content false

Crawling an Entire Site

firecrawl.js crawl --url "https://example.com" --limit 50

When: You need content from multiple pages across a site. This is an async operation — the CLI starts a crawl job and returns a job ID.

Options:

• --limit N — Maximum number of pages to crawl
• --max-depth N — Maximum link depth from the starting URL
• --poll — Automatically poll for completion instead of returning the job ID
• --dry-run — Preview the API request without sending it

Workflow without --poll:

1. Start the crawl: firecrawl.js crawl --url "https://example.com" --limit 20
2. Note the returned job ID
3. Check status: firecrawl.js crawl-status --id "job-id-here"
4. Repeat step 3 until complete

Workflow with --poll (recommended):

firecrawl.js crawl --url "https://example.com" --limit 20 --poll

The CLI handles polling automatically and returns all results when done.

Checking Crawl Status

firecrawl.js crawl-status --id "job-id-here"

When: You started a crawl without --poll and need to check if it finished.

Getting a Site Map

firecrawl.js map --url "https://example.com"

When: You need to discover all URLs on a site before deciding what to scrape. Good first step before targeted scraping.

Options:

• --search "term" — Filter URLs containing a search term
• --limit N — Maximum number of URLs to return
• --dry-run — Preview the API request without sending it

Example — find all blog pages:

firecrawl.js map --url "https://example.com" --search "blog"

Searching and Scraping

firecrawl.js search --query "site:example.com pricing plans"

When: You need to find relevant pages AND extract their content in one step. Combines web search with content extraction.

Options:

• --limit N — Number of results to return
• --country "us" — Country code for localized results
• --dry-run — Preview the API request without sending it

Example — find competitor pricing pages:

firecrawl.js search --query "competitor.com pricing" --limit 5

Autonomous Agent Extraction

firecrawl.js agent --prompt "Find pricing information for Acme Corp" --poll

When: You need the AI to autonomously search, navigate, and gather data. No URLs required -- just describe what you need.

Options:

• --prompt "..." -- Describe what data to find (required, max 10000 chars)
• --url "..." -- Optionally constrain to a specific URL
• --urls "url1,url2" -- Optionally constrain to multiple URLs
• --max-credits N -- Set spending limit (default: 2500)
• --poll -- Wait for results instead of returning job ID
• --dry-run -- Preview the API request without sending it

Example -- find competitor pricing:

firecrawl.js agent --prompt "Find pricing tiers and plans for competitor.com" --url "https://competitor.com" --poll

Extracting Structured Data

firecrawl.js extract --urls "https://example.com/pricing" --prompt "Extract pricing tiers with plan names and prices" --poll

When: You need specific structured data extracted from URLs using LLMs. Supports wildcards for domain-wide extraction.

Options:

• --urls "url1,url2" -- URLs to extract from, comma-separated (required). Supports /* wildcards
• --prompt "..." -- Describe what data to extract
• --schema '{"type":"object",...}' -- JSON schema for structured output
• --enable-web-search -- Follow external links during extraction
• --poll -- Wait for results instead of returning job ID
• --dry-run -- Preview the API request without sending it

Example -- extract product data from multiple pages:

firecrawl.js extract --urls "https://example.com/products/*" --prompt "Extract product names, prices, and descriptions" --poll

Choosing the Right Command

• Need one page? → scrape
• Need the whole site? → crawl with --limit and --poll
• Need to discover URLs first? → map, then scrape specific pages
• Need search + content? → search
• Started a crawl and need results? → crawl-status
• Need AI to find data autonomously? → agent with --poll
• Need structured data from known URLs? → extract with --prompt and --poll

Common multi-step workflows:

1. Targeted extraction: map → review URLs → scrape specific pages
2. Full site dump: crawl --poll --limit 100
3. Research a topic: search → review results → scrape for deeper content

Output Format

Present extracted content with clear structure:

• Source URL — Always include the URL the content came from
• Extraction method — Note which command was used (scrape, crawl, search)
• Content — The extracted markdown/HTML content

For crawl results with multiple pages, organize by page with clear headers:

### Page 1: [title] (https://example.com/page-1)
[content]

### Page 2: [title] (https://example.com/page-2)
[content]

Environment Setup

The Firecrawl CLI requires a FIRECRAWL_API_KEY environment variable. If the key is not set, the CLI will return an error. The user needs to:

1. Get an API key from firecrawl.dev
2. Set it in their environment: export FIRECRAWL_API_KEY="fc-..."

Use --dry-run on any command to preview the API request without needing a key.

Related Skills

• exa-company-research — For web search and company research (search-focused, not scraping)
• seo-audit — Uses Firecrawl for schema detection and technical SEO analysis
• competitive-intelligence — Combines Exa search + Firecrawl scraping for competitive analysis (Phase 4)