What is Firecrawl?
Firecrawl is a search, scraping, and web interaction API that returns clean content in Markdown, structured JSON, HTML, or screenshots for input into language models, RAG, or AI agents. The project, according to GitHub (as read on 25/10/2026), is primarily under the AGPL-3.0 license (SDKs and some interface components are MIT), has around 195 thousand stars, and offers both self-hosted and cloud service options. According to the official pricing page, the Free plan includes 1,000 credits per month, the Hobby plan is $19 per month for 5,000 credits, and each scraped page costs 1 credit.
The article covers: 1) endpoints; 2) pricing table and how credits are deducted; 3) when to use it; 4) legal and ethical considerations when scraping; 5) OPCAIOS experience. For example, a team building RAG could scrape 5,000 public documentation pages using Firecrawl, exhaust the 5,000 credits from the Hobby plan, then feed the Markdown into a knowledge base.
Firecrawl transforms websites into clean data for AI; credits are counted per page. Diagram: OPCAIOS.
What can Firecrawl do through its main endpoints?
Firecrawl is a context API for AI agents, as the project introduces itself on GitHub (as read on 25/10/2026). The main endpoints include:
- Scrape: converts a URL into Markdown, HTML, screenshot, or structured JSON.
- Search: finds the web and returns full content of results.
- Crawl: scrapes every page of a site and returns a job code to check status.
- Map: lists a site's URLs, can filter by relevance level.
- Interact: runs AI prompts or code on a scraped page, like clicking or filling forms.
- Agent: natural language descriptions to automatically collect data, replacing the old
/extract. - Batch Scrape, Parse, Monitor: async batch scraping, reading PDFs and DOCX, alerting when pages change.
The documentation states Firecrawl handles JavaScript-heavy pages, dynamic content, and anti-bot measures via proxy. The service has SDKs for many languages like Python, Node.js, Go, the MCP server firecrawl-mcp, and integrations with Lovable, Zapier, n8n. The 96% web coverage and 3.4-second latency (P95) are self-reported by the provider; we have not verified them. See also What is RAG? and What is MCP?.
What pricing plans does Firecrawl offer?
Firecrawl has six plans, and according to the official pricing page (as read on 25/10/2026, USD, before tax), pricing is monthly as follows.
- Free: $0, no card required, 1,000 credits and 2 concurrent requests.
- Hobby: $19 (or $16 if paid annually), 5,000 credits and 5 concurrent requests.
- Standard: $99 (or $83 if paid annually), 100,000 credits and 25 requests.
- Growth: $399 (or $333 if paid annually), 500,000 credits and 50 requests.
- Scale: $749 (or $599 if paid annually), 1,000,000 credits and 100 requests.
- Enterprise: negotiated, customized.
For example, a team that only needs to scrape a few thousand pages monthly should start with Hobby, then upgrade to Standard when the site grows to hundreds of thousands of pages. Ultimately, search and scraping share a credit pool: the Hobby plan is enough for 2,500 searches or 5,000 scraped pages. Pricing may change; confirm on firecrawl.dev.
Detailed pricing table
According to the official pricing page (USD, as read on 25/10/2026, before tax):
| Plan | Price per month (paid annually) | Credits per month | Concurrent requests |
|---|---|---|---|
| Free | 0 (no card required) | 1,000 | 2 |
| Hobby | 19 (16) | 5,000 | 5 |
| Standard | 99 (83) | 100,000 | 25 |
| Growth | 399 (333) | 500,000 | 50 |
| Scale | 749 (599) | 1,000,000 | 100 |
| Enterprise | Negotiated | Customized | Customized |
When should you use Firecrawl?
| Situation | Recommendation |
|---|---|
| Need clean page content to feed AI (RAG, summarization, competitor tracking) | Firecrawl is suitable |
| Only need a few pages, occasionally | Built-in AI assistant page reader or manual download |
| Need to click, log in, perform multi-step actions | Browser agent; see What is Browser Use? |
| Need data with an official API | Use the API (see What is API?) |
| Want full control and don't want to send data outside | Self-host the open-source codebase, plus server and maintenance costs |
What to Watch Out For When Web Scraping with Firecrawl: Licensing and Legal Issues
Web scraping with Firecrawl means downloading other people's content, so there are five things to watch out for according to GitHub (read 25/10/2026) and legal practice.
- AGPL-3.0: if you self-host and modify then provide the service over a network, the license comes with an obligation to share source code; read carefully or ask an expert before using commercially (the MIT SDK is lighter).
- robots.txt and site terms: Firecrawl respects robots.txt by default, and the code repository states that users are responsible for complying with website policies. Many sites ban scraping in their terms.
- Copyright and personal data: scraped content has copyright; personal data needs a legal basis (ask a legal expert). Don't use other people's content to copy into your own posts.
- Prompt injection: scraped pages may contain harmful instructions targeting AI; use the prompt injection check option (adds 4 credits per page) and treat scraped content as untrustworthy.
- Cost: credits are counted per page; scraping large sites drains credits fast, so use Map first to estimate.
OPCAIOS Experience
We haven't used Firecrawl; our Vault only has this tool's name in reference records about the open source ecosystem. When reading sources while writing posts, we use built-in page readers and integrated browsers, and clearly note in articles which pages we couldn't read instead of finding ways around blocks. Our system also has a competitor tracking function using public services. Since we haven't tried Firecrawl, we don't claim anything about its Markdown quality or success rate.
Read more: 5 Best Web Data Scraping Tools for AI in 2026.
Frequently Asked Questions
What is Firecrawl?
It's a web scraping and search API that returns clean content (Markdown, JSON) for AI, with an open source version and a cloud service.
Is Firecrawl free?
Firecrawl has a Free plan with 1,000 credits per month and 2 concurrent requests, no card needed, according to the official pricing page (read 25/10/2026). There are three ways to use it, from free to paid. First, the Free plan is enough to try a few hundred pages and measure Markdown quality. Next, the Hobby plan costs $19 per month ($16 if you pay yearly) for 5,000 credits, enough to scrape 5,000 pages in Markdown format. Finally, the open source version can be self-hosted without paying for the service, but according to GitHub the project is mainly under AGPL-3.0 and you pay for server, maintenance yourself. For example, a newcomer can scrape 200 pages using the Free plan to check, then decide whether to upgrade. Pricing doesn't include tax and may change, so confirm on the official site before buying.
How many credits does it cost to scrape one page?
The cost to scrape one page is 1 credit, according to the credit table on Firecrawl's official pricing page (read 25/10/2026). The credit deduction rules are:
- Scrape: 1 credit per page; getting structured JSON adds 4 credits per page; turning on prompt injection check adds 4 credits per page.
- Crawl: 1 credit per page.
- Search: 2 credits for 10 results.
- Map: 1 credit per call.
- PDF: 1 credit per page.
First, scraping 100 pages with Firecrawl in Markdown format costs 100 credits, while getting structured JSON for the same 100 pages costs 500 credits. With the Hobby plan at 5,000 credits ($19 per month), this level equals 1,000 JSON pages. Finally, since scraping large sites drains credits fast, use Map first to estimate the page count.
Is web scraping legal?
It depends on the content, site terms, and the law; respect robots.txt, copyright, personal data, and ask a legal expert.
Should I self-host or use the cloud?
Cloud is more convenient (from $19 per month for Hobby according to the official pricing page); self-hosting gives you control but comes with AGPL-3.0, servers, and maintenance. Want to learn how to feed web data into AI workflows? Check out OPCAIOS's AI Application Training course.
Source: firecrawl/firecrawl GitHub repository, firecrawl.dev/pricing and docs.firecrawl.dev, read 25/10/2026; stars and performance metrics are published by the project, unverified; pricing doesn't include tax and may change. The experience is OPCAIOS's.
Translated from Vietnamese with AI. Read the original.
Want to apply this in your business?
Book a free 30-minute call (Vietnam time, in English or Vietnamese).