To use Firecrawl Search in OpenClaw, run `clawhub install firecrawl-search`, get a free API key at firecrawl.dev, and set FIRECRAWL_API_KEY in your OpenClaw config. Unlike search skills that return URLs and snippets, Firecrawl Search extracts the full structured content from matched pages — turning search queries into ready-to-use data, not just a list of links to read manually.
Beyond Links: Firecrawl Search Returns the Full Page, Not Just the URL
Most search skills in OpenClaw answer the question 'where can I find this information?' — they return a list of URLs with short snippets so you can decide which pages to visit. Firecrawl Search answers a different question: 'what does this content actually say?' It searches for relevant pages and then immediately fetches and extracts the full text content from those pages, returning cleaned Markdown ready for OpenClaw to process, summarize, or analyze without any further steps.
This distinction matters enormously for automation. If you want OpenClaw to monitor a news topic and summarize the key developments, Tavily AI Search gives you titles and URLs — which are useful for reading, but require additional fetching to get the actual content. Firecrawl Search gives you the full article text directly. For pipelines that need to extract specific data points from web content (prices, names, dates, statistics), Firecrawl's structured output is far more useful than a snippet that may not include the specific data you are targeting.
The companion skill, firecrawl-skills, provides the full Firecrawl toolkit including multi-page crawling, site mapping, and structured data extraction with custom schemas. Firecrawl Search is the focused version: search + extract, one query at a time. For most OpenClaw users who need richer search results without the complexity of full site crawling, Firecrawl Search is the right starting point — and the same FIRECRAWL_API_KEY works for both skills if you later decide to expand to the broader toolkit.
Integration method
Firecrawl Search installs as a ClawHub skill and uses FIRECRAWL_API_KEY to call Firecrawl's search and scrape API. What distinguishes this skill from Tavily or Brave Search is the output: instead of returning a list of URLs with short snippets, Firecrawl fetches the matched pages, strips ads and navigation, extracts the main content, and returns it in clean Markdown. You get full article text, structured data, and complete page content — not just enough to decide which link to click.
Prerequisites
- OpenClaw installed and running on your system
- ClawHub CLI available in your terminal
- A Firecrawl account at firecrawl.dev for your API key
- Awareness that Firecrawl Search returns more content per query than snippet-based search skills, so responses may be longer
Step-by-step guide
Install the Firecrawl Search Skill
Install the Firecrawl Search Skill
Install the Firecrawl Search ClawHub skill using the exact package name `firecrawl-search`. This is the targeted search + extract skill. The related `firecrawl-skills` package provides the full Firecrawl toolkit with multi-page crawling and site mapping — if you only need search-based extraction, `firecrawl-search` is the right starting point. Both skills use the same FIRECRAWL_API_KEY, so if you install both at any point, you only need to configure the API key once. OpenClaw reads the single key for both skill packages. The installation process registers the skill with your local OpenClaw configuration and validates the required dependencies. After installation, run `clawhub info firecrawl-search` to see all available configuration parameters before setting up your API key.
1# Install Firecrawl Search skill2clawhub install firecrawl-search34# Verify installation5clawhub list | grep firecrawl67# See available configuration options8clawhub info firecrawl-searchPro tip: If you think you might need full site crawling later, run `clawhub install firecrawl-skills` at the same time — both use the same API key and you will not need to reconfigure anything when you add the second skill.
Expected result: Terminal confirms 'firecrawl-search@X.X.X installed successfully. Required config: FIRECRAWL_API_KEY'. The skill appears in `clawhub list`.
Get Your Firecrawl API Key
Get Your Firecrawl API Key
Go to firecrawl.dev and sign up for a free account. Firecrawl's free tier provides a limited number of credits per month — each page crawled (fetched and extracted) consumes credits. Navigate to your Firecrawl dashboard after signing up and look for the 'API Keys' section. Click 'Create new key' or look for your default API key on the dashboard home. The key starts with `fc-` followed by a long alphanumeric string. Copy it immediately. Firecrawl's credit system is worth understanding before you start: each search + extract operation in OpenClaw typically consumes 5-10 credits depending on the number of pages extracted and the content volume. The free tier provides enough credits to experiment and for light personal use. Paid plans scale from $16/month for 3,000 credits to higher tiers for production use. Note that Firecrawl Search is a more credit-intensive operation than Tavily or Brave Search because it does not just return snippets — it actually fetches and extracts full pages. Plan your usage accordingly.
Pro tip: The Firecrawl dashboard shows your remaining credits in real time. For automated OpenClaw workflows that use Firecrawl Search frequently, set up a usage alert in the Firecrawl dashboard to notify you when credits fall below a threshold.
Expected result: You have a Firecrawl API key starting with `fc-` from the Firecrawl dashboard, ready to configure in OpenClaw.
Configure FIRECRAWL_API_KEY in OpenClaw
Configure FIRECRAWL_API_KEY in OpenClaw
Set your Firecrawl API key in OpenClaw's configuration under the variable name FIRECRAWL_API_KEY. This variable name is shared by both the `firecrawl-search` and `firecrawl-skills` ClawHub packages — if you install both, one configuration entry covers both. Use the clawhub config command to set the key, or add it to your ~/.openclaw/.env file. After configuring, reload OpenClaw to activate the skill. Run a quick verification to confirm the key is stored correctly.
1# Set Firecrawl API key2clawhub config set FIRECRAWL_API_KEY fc-your-api-key-here34# Or add to ~/.openclaw/.env5# ~/.openclaw/.env6FIRECRAWL_API_KEY=fc-your-api-key-here78# Reload and verify9clawhub reload10clawhub config get FIRECRAWL_API_KEYPro tip: FIRECRAWL_API_KEY works for both firecrawl-search and firecrawl-skills if you have both installed. You only need to configure this key once regardless of how many Firecrawl skills you add.
Expected result: `clawhub config get FIRECRAWL_API_KEY` returns a masked key starting with fc-. OpenClaw confirms firecrawl-search is active and authenticated.
Test Full Content Extraction in OpenClaw Chat
Test Full Content Extraction in OpenClaw Chat
Open OpenClaw chat and run a query that demonstrates Firecrawl's full-content extraction. The response will be noticeably different from what you see with Tavily or Brave — instead of a list of URLs with 150-character snippets, you will receive the actual page content from the matched pages, cleaned and formatted as Markdown. For your first test, ask OpenClaw to search and extract something with clearly structured content — a product page, a technical documentation article, or a news piece. This makes it easy to verify that Firecrawl is returning full text rather than snippets. Note that Firecrawl Search responses are longer than standard search results. If you are using OpenClaw in a context with message length limits, consider configuring a lower `limit` value (number of pages extracted per query) to keep response sizes manageable.
Search for the official Stripe pricing page and extract the complete pricing information including all plan tiers, per-transaction fees, and any volume discount information listed on the page.
Paste this in OpenClaw chat
Pro tip: If OpenClaw's response is truncated for long Firecrawl results, ask it to summarize or extract only specific data points rather than returning all extracted content verbatim. Firecrawl does the heavy lifting of extraction — OpenClaw processes the content to answer your specific question.
Expected result: OpenClaw returns the full extracted text from the matched pages in Markdown format, including the complete article or page content — not just a title and snippet. You can read the actual content or ask OpenClaw to analyze it further.
Configure for Data Extraction Pipelines
Configure for Data Extraction Pipelines
For production use cases — automated monitoring, competitive intelligence, or data collection pipelines — configure Firecrawl Search with appropriate limits and output formats. The key settings are `limit` (how many pages to extract per query) and `scrapeOptions` (what page elements to include or exclude). Firecrawl's extraction strips JavaScript, navigation menus, footers, ads, and boilerplate by default, returning clean main content. For pages with important data in tables or structured formats, the Markdown output preserves these structures accurately. RapidDev can help teams build production-grade Firecrawl Search pipelines in OpenClaw — particularly for ongoing competitive monitoring or research automation workflows that need reliable, structured data extraction from the web at scale.
Search for the top 3 results about 'OpenAI API rate limits 2026' and extract the complete technical documentation from each page. Present the specific rate limit numbers in a structured format.
Paste this in OpenClaw chat
1# ~/.openclaw/skills/firecrawl-search.yaml2firecrawl-search:3 limit: 5 # max pages to extract per search (default: 5)4 lang: en # language preference for results5 country: us # country code for result localization6 tbs: "" # time-based search filter (qdr:d=past day, qdr:w=past week)7 scrapeOptions:8 formats:9 - markdown # return clean markdown (recommended)10 # - html # raw HTML (use only if you need structure)11 onlyMainContent: true # strip nav, ads, footers (default: true)12 includeTags: [] # optional: only include specific HTML tags13 excludeTags: # optional: exclude specific elements14 - nav15 - footer16 - asidePro tip: Keep `onlyMainContent: true` (the default) for article and documentation extraction — it significantly improves content quality by stripping boilerplate. Only set it to false if you need the full page HTML structure for a specific extraction task.
Expected result: Configured extraction returns clean Markdown content from matched pages with navigation and ads stripped. The `limit` setting controls how many pages are fully extracted per search query.
Common use cases
Full-Text News and Content Extraction
Search for articles on a topic and get the complete article text back — not just headlines. Use this to have OpenClaw read and analyze actual content rather than deciding which links to follow manually. Particularly useful for daily briefings where you want OpenClaw to digest full articles and produce a summary.
Search for the 5 most recent articles about enterprise AI adoption challenges published this week. Extract the full text of each article and give me a comprehensive summary of the main themes and specific data points mentioned.
Copy this prompt to try it in OpenClaw
Pricing and Product Data Extraction
Search for competitor pricing pages or product listings and extract the structured data from those pages. Firecrawl's extraction strips navigation and ads, returning just the relevant pricing or product information in a format OpenClaw can parse and compare.
Search for the pricing pages of the top 5 project management tools (Asana, Linear, Notion, Monday, ClickUp) and extract their current pricing tiers, feature lists, and per-seat costs. Present the data in a comparison table.
Copy this prompt to try it in OpenClaw
Research Synthesis From Multiple Full Sources
Collect full content from several relevant pages on a research topic and synthesize a comprehensive answer. Rather than reading each source yourself, Firecrawl extracts the content and OpenClaw synthesizes it into a unified analysis.
Search for detailed technical explanations of how Retrieval-Augmented Generation works and find the 3 most thorough sources. Extract the full technical content from each and synthesize a comprehensive technical overview.
Copy this prompt to try it in OpenClaw
Troubleshooting
OpenClaw shows 'firecrawl-search: authentication failed' or 'Invalid API key'
Cause: FIRECRAWL_API_KEY is not set in OpenClaw's configuration, was set under a different variable name, or the key was entered without the `fc-` prefix.
Solution: Run `clawhub config get FIRECRAWL_API_KEY` to verify the key is set and starts with `fc-`. If empty or incorrect, re-run `clawhub config set FIRECRAWL_API_KEY fc-your-key`. Verify the key in the Firecrawl dashboard — keys can be regenerated if lost.
1# Check and reset2clawhub config get FIRECRAWL_API_KEY3clawhub config set FIRECRAWL_API_KEY fc-XXXXXXXXXXXXXXXXXXXX4clawhub reloadFirecrawl Search returns 'Insufficient credits' or stops extracting after a few queries
Cause: The Firecrawl free tier credit allowance has been exhausted for the month. Each page extraction consumes credits — high-volume use depletes the free tier quickly.
Solution: Check your remaining credits in the Firecrawl dashboard at firecrawl.dev. Upgrade to a paid plan if you need sustained usage. To preserve credits, reduce the `limit` config to 2-3 pages per query instead of the default 5, and use Firecrawl Search only when full content extraction is necessary — route general queries to Tavily or Brave Search instead.
1# Reduce pages extracted per query to preserve credits2firecrawl-search:3 limit: 2 # extract only 2 pages instead of default 5Extracted content includes navigation menus, footers, and unrelated boilerplate text
Cause: The `onlyMainContent` setting is disabled, causing Firecrawl to return the full raw page content rather than the extracted main article text.
Solution: Ensure `onlyMainContent: true` is set in your firecrawl-search.yaml configuration file. This is the default but may have been changed. Reload OpenClaw after updating the config.
1# Ensure main content extraction is enabled2firecrawl-search:3 scrapeOptions:4 onlyMainContent: trueClawHub install firecrawl-search installs but the skill does not appear active in OpenClaw
Cause: OpenClaw was not reloaded after installation, or the skill requires a configuration key that has not been set — it may install but stay inactive until FIRECRAWL_API_KEY is configured.
Solution: Set FIRECRAWL_API_KEY and then run `clawhub reload` to force OpenClaw to re-read the configuration and activate all properly configured skills. Run `clawhub status` to see which skills are active versus installed-but-inactive.
1# Set key and reload2clawhub config set FIRECRAWL_API_KEY fc-your-key3clawhub reload4clawhub status | grep firecrawlBest practices
- Use Firecrawl Search when you need the full page content, not just a snippet to decide which link to click — for research that requires reading and analyzing actual article text rather than scanning headlines.
- Route general queries to Tavily AI Search and reserve Firecrawl Search for queries where full content extraction adds clear value — this preserves your Firecrawl credits for high-value use cases.
- Set `limit: 2` or `limit: 3` in your config for exploratory use — extracting 5 pages (the default) consumes significantly more credits than extracting 2-3 when you are testing or doing light research.
- Keep FIRECRAWL_API_KEY in ~/.openclaw/.env and never in chat history or code repositories — Firecrawl API keys provide access to your paid credit balance.
- Monitor Firecrawl credit consumption in the Firecrawl dashboard weekly when running automated pipelines — set up usage alerts to avoid unexpected credit depletion during high-activity periods.
- For structured data extraction (prices, tables, lists), tell OpenClaw explicitly what data to extract from the Firecrawl content — the extraction is most valuable when paired with a specific parsing task rather than returning raw text.
- Combine firecrawl-search with firecrawl-skills for comprehensive extraction workflows — use firecrawl-search to identify relevant pages by topic, then use firecrawl-skills for deep crawling of the sites those pages belong to.
- Test extraction quality on a few representative pages before running a bulk pipeline — some sites use techniques (aggressive JavaScript rendering, paywalls, bot detection) that affect Firecrawl's extraction quality.
Alternatives
Firecrawl Skills provides the full Firecrawl toolkit including multi-page crawling and site mapping — better when you need to extract data from entire websites rather than searching for and extracting individual pages.
Deep Scraper recursively follows links to build comprehensive datasets from websites — better for deep extraction from a single known site rather than Firecrawl Search's approach of finding and extracting relevant pages across the open web.
Tavily AI Search returns AI-ranked snippets and synthesized answers without full page extraction — better for quick research questions where you need a curated answer rather than the raw page content.
Web Search Plus aggregates results from multiple search backends for broader coverage — better when you need wide-net search results rather than Firecrawl's deep extraction of fewer pages.
Frequently asked questions
How do I install Firecrawl Search in OpenClaw?
Run `clawhub install firecrawl-search` in your terminal. Get a free API key from firecrawl.dev, then configure it with `clawhub config set FIRECRAWL_API_KEY fc-your-key`. Run `clawhub reload` and you can immediately use structured web extraction in OpenClaw chat.
What is the difference between Firecrawl Search and Tavily AI Search in OpenClaw?
Tavily AI Search returns snippets, summaries, and ranked URLs — useful for finding where information is and getting a synthesized overview. Firecrawl Search fetches the actual full page content and returns it as clean Markdown — useful when you need to read, analyze, or extract specific data from the pages themselves. Firecrawl is more credit-intensive but returns far more usable content per query.
What is the FIRECRAWL_API_KEY and where do I get it?
FIRECRAWL_API_KEY is the OpenClaw environment variable for authenticating with Firecrawl's API. Get your key from the Firecrawl dashboard at firecrawl.dev after signing up. The key starts with `fc-`. The same key works for both firecrawl-search and firecrawl-skills if you install both.
How many searches can I do with Firecrawl in OpenClaw for free?
Firecrawl uses a credit system rather than a fixed request count. Each page extracted consumes credits, and a search query that extracts 5 pages uses roughly 5x more credits than a single-page extraction. The free tier provides a limited monthly credit allowance — check firecrawl.dev/pricing for current allowances, as these change. Paid plans start at $16/month for 3,000 credits.
What is the difference between firecrawl-search and firecrawl-skills in OpenClaw?
firecrawl-search is the focused skill for search queries — it finds relevant pages and extracts their content in one step. firecrawl-skills is the broader toolkit that includes multi-page site crawling, site mapping, and custom schema extraction for structured data pipelines. Both use the same FIRECRAWL_API_KEY. Start with firecrawl-search for typical research needs and add firecrawl-skills if you need full-site crawling capabilities.
Can RapidDev help with Firecrawl Search pipelines in OpenClaw?
Yes — RapidDev can help configure Firecrawl Search as part of automated extraction and monitoring pipelines in OpenClaw, particularly for competitive intelligence workflows or content aggregation systems that need structured data extraction at scale. The self-serve setup here covers standard individual use; teams with production pipeline requirements can contact RapidDev for specialized support.
Why is Firecrawl Search slower than other search skills in OpenClaw?
Firecrawl Search is inherently slower because it does not just return search result metadata — it actually fetches and processes each matched page to extract the full content. This additional step adds 3-10 seconds compared to snippet-based skills. The trade-off is significantly more useful output per query. For time-sensitive applications, use Tavily or Brave Search instead.
Talk to an Expert
Our team has built 600+ apps. Get personalized help with your project.
Book a free consultation