Firecrawl: The Open-Source Web Data API for AI, Explained
Firecrawl is a web data API for AI. Give it a URL and it returns the page as clean markdown or structured JSON, with navigation, footers and ads stripped out, ready for a language model. It also searches the web and returns full page content, crawls entire sites, maps their structure, reads PDFs and Word files, and can click, type and scroll through pages that need interaction. It’s open source, with about 201,000 stars on GitHub, which makes it one of the most popular developer tools in AI. In September 2026, it raised a $75 million Series B and launched Alexandria, a library of licensed data for agents.
The quick answer: Firecrawl is the go-to tool for turning websites into data an AI can use, especially when you know which sites you need. It’s easy to start, with a free tier, a one-command setup for coding agents, and connectors for ChatGPT and Claude, and you can self-host it. Its search has grown into a real alternative to search-first APIs, though scraping remains its core strength. Everything below comes from Firecrawl’s site, pricing, developer summary, GitHub repository, Y Combinator profile and press coverage.
Firecrawl at a glance
Checked against Firecrawl’s site, pricing, developer summary and GitHub repository on November 13, 2026.
| Firecrawl | |
|---|---|
| What it is | An API that searches, scrapes, crawls and interacts with the web for AI |
| Endpoints | Search, Scrape, Crawl, Map, Parse, Interact, Agent and Monitor |
| Output | Markdown, structured JSON, screenshots and more |
| Open source | AGPL-3.0, about 201,000 GitHub stars; self-hostable |
| Works with | ChatGPT plugin, Claude connector, MCP server, CLI for Cursor, Claude Code and Windsurf |
| Pricing | Free 1,000 credits a month; paid plans from $16 a month billed annually |
| Company | San Francisco; founded by Caleb Peffer (CEO), Eric Ciarla and Nicolas Silberstein Camara; Y Combinator S22 |
What Firecrawl does
Firecrawl’s developer summary calls it “the open-source web data API for AI agents and LLM applications.” Its homepage puts it more simply: “The infrastructure layer that helps AI find, read, and act on the live web.”
The endpoints cover each step an agent might need:
| Endpoint | What it does |
|---|---|
| Search | Searches the web and returns full page content, not just links or snippets |
| Scrape | Turns a page, including JavaScript-heavy ones, into clean markdown or structured JSON |
| Crawl | Scrapes every subpage of a site |
| Map | Discovers a site’s structure and pages |
| Parse | Extracts text from PDFs, Word files and other documents |
| Interact | Clicks, types, fills forms and scrolls through pages that need it |
| Agent | Takes a description of the data you need and finds it, no URLs required |
| Monitor | Checks pages, crawls or searches on a schedule and alerts you to changes by webhook, email or Slack |
A few practical details make Firecrawl easy to rely on. It’s a managed service, so there’s no browser to run yourself. A “smart wait” feature lets slow-loading pages finish before scraping, and Firecrawl doesn’t charge for scrapes that return nothing. The company says its cleaned markdown uses 93% fewer input tokens than raw pages, and on its own 1,000-URL benchmark from January 2026 it reports 96% coverage with a 95th-percentile response time of about 3.4 seconds.
Alexandria: licensed data for agents
Alongside its Series B in September 2026, Firecrawl launched Alexandria, a library of data providers and specialized indexes for companies, people, finance, legal, real estate and more, plus curated datasets such as tens of millions of scientific paper abstracts and code files. Agents call it through the same API, MCP server or CLI.
Alexandria is also Firecrawl’s answer to a growing tension around AI and the web. The homepage describes “Fair access to web content, licensed through Firecrawl Alexandria,” and SiliconANGLE reported that the company plans a self-service content licensing system. Firecrawl’s guidance to agents is to check Alexandria first for data a provider already publishes, like SEC filings or company records, before scraping.
How developers connect
Firecrawl is unusually easy to plug in. In ChatGPT it’s a plugin, in Claude it’s a connector, and coding agents add it as an MCP server. A single command, npx -y firecrawl-cli@latest init --all --browser, gives Cursor, Claude Code, Windsurf and other tools web search and scraping. Agents without an API key can search, scrape and interact on a limited free tier, and there are SDKs for Python and other languages plus integrations with most agent frameworks.
Self-hosting and the license
Firecrawl’s code is open source under the AGPL-3.0 license, and teams can run it on their own servers. AGPL is a “copyleft” license: if you modify Firecrawl and offer it to others over a network, you’re required to share your changes. That’s fine for most internal use, but companies building a commercial product on a modified version should check with their lawyers or use the hosted API. One developer quoted on Firecrawl’s pricing page wrote of self-hosting it, “There’s no crawl job I haven’t done with it.”
Pricing
Firecrawl prices in credits shared across every endpoint. A basic scrape or crawled page costs 1 credit, a Map call 1 credit, 10 search results 2 credits, and a minute of browser interaction 2 credits. Structured JSON output adds 4 credits per page.
| Plan | Monthly credits | Monthly billing | Annual billing (per month) |
|---|---|---|---|
| Free | 1,000 | $0 | $0 |
| Hobby | 5,000 | $19 | $16 |
| Standard | 100,000 | $99 | $83 |
| Growth | 500,000 | $399 | $333 |
| Scale | 1,000,000 | $749 | $599 |
| Enterprise | Custom | Custom | Custom |
Higher plans also run more browsers at once, from 2 on the free plan to 100 on Scale. Paid plans can turn on pay-as-you-go usage above their credits, up to a limit you set.
The company
Firecrawl is based in San Francisco and was founded by Caleb Peffer, its CEO, Eric Ciarla and Nicolas Silberstein Camara. Its Y Combinator profile lists the Summer 2022 batch. In August 2025 it raised a $14.5 million Series A led by Nexus Venture Partners, with Y Combinator and Shopify CEO Tobias Lütke participating. In September 2026, it closed a $75 million Series B led by Smash Ventures, joined by Y Combinator, Altos Ventures, Nexus Venture Partners, Freestyle and Offline Ventures, with much of the money going to adding more data providers to Alexandria.
Firecrawl’s GitHub repository, created in April 2024, has about 201,000 stars and more than 10,500 forks. We didn’t find Firecrawl on independent review sites; feedback lives mostly in developer communities and on social media.
How Firecrawl compares
Firecrawl overlaps with the search APIs Exa, Tavily, Parallel and the Brave Search API, all of which now offer some content extraction. The difference is emphasis. Exa and Parallel run their own search indexes and focus on finding the right pages and doing research; Tavily focuses on fast setup and safety filtering for agent search. Firecrawl’s strength is reading and acting on specific sites thoroughly, including crawling, documents and interaction, and it’s the only one in the group you can fully self-host. Context.dev, a newer scraping API, competes with Firecrawl head-on and adds company brand data. For browser automation beyond Firecrawl’s Interact, teams look at dedicated browser platforms.
If you’d rather not build any of this, Carly searches the web as part of its work and connects to thousands of apps, so from ChatGPT, Claude or a text message you can ask it to pull details from a company’s website and log them in your CRM or a spreadsheet.
Who Firecrawl suits
Firecrawl is a good fit if:
- You need clean content from specific websites, documentation or PDFs for an AI.
- You want to crawl whole sites or monitor pages for changes.
- You want an open-source tool you can self-host.
- You want web tools in Claude Code, Cursor or Codex with one command.
Look elsewhere if:
- Your main need is ranking the best sources for open-ended questions; a search-first API may fit better.
- You need long-running, persistent browser sessions; Firecrawl says Interact isn’t designed for them.
- You want a finished assistant rather than an API.
FAQ
What is Firecrawl?
Firecrawl is an open-source web data API for AI. It searches the web, scrapes pages into clean markdown or JSON, crawls sites, parses documents and can interact with pages, so AI agents can use web content.
Is Firecrawl free?
Firecrawl’s free plan includes 1,000 credits a month with no credit card, and the open-source version can be self-hosted. Paid plans start at $19 a month, or $16 a month billed annually.
Is Firecrawl open source?
Yes. Firecrawl’s code is licensed AGPL-3.0 and has about 201,000 GitHub stars. Teams can self-host it or use the managed API.
Does Firecrawl work with Claude and ChatGPT?
Yes. Firecrawl is available as a Claude connector, a ChatGPT plugin and an MCP server for coding agents like Claude Code and Cursor.
Who founded Firecrawl?
Caleb Peffer, Eric Ciarla and Nicolas Silberstein Camara founded Firecrawl. It raised a $75 million Series B led by Smash Ventures in September 2026.
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
