There is a genuine free tier, and it renews: 1,000 credits a month — roughly 500 searches or 1,000 pages — with two concurrent requests and no card. Paid plans start at $16 a month billed yearly for 5,000 credits and 5 concurrent requests, and run to $599 for a million credits and 100 concurrent requests. Enterprise is quoted.

The unit everything here is metered in. A basic scrape costs 1. A 2026 review records that Stealth Mode and AI-schema /extract calls cost up to 5, which is why a plan sized on simple pages runs short once you start on sites that resist scraping.

Can I self-host Firecrawl?

Yes. The core is AGPL-3.0, the SDKs are MIT, and self-hosting is supported. What you do not get is parity: a GitHub issue on the project records that the self-hosted build lacks the managed anti-bot "fire engine" and stealth proxying the cloud product has. AGPL also means that running a modified copy as a network service requires publishing your modifications.

Firecrawl or Browser Use?

Read a page versus drive a browser. Firecrawl fetches a URL and returns clean text, with no browser session involved. Browser Use puts an agent in a real browser to click, type and log in. If all you need is what the page says, Firecrawl is the simpler and more predictable half; Browser Use earns its keep when the task needs a form filled or a login completed.

Is the project active?

177,312 GitHub stars and a push on 6 September 2026. A $14.5M Series A led by Nexus Venture Partners closed in August 2025, and the company claims 350,000+ developers and 15x year-on-year growth — its own figures. Recent ships: a Developer Index over 70M code artifacts on 20 August 2026, and a Life Sciences Index over 41M papers on 13 August 2026.

Firecrawl explained: pricing, limits, alternatives (2026) • SaikoDev

Good for / not for

Anyone feeding web content to an AI: RAG, agents, research pipelines.
Starting without a card, on an allowance that renews — 1,000 credits a month, about 500 searches or 1,000 pages.
Anyone weighing project health: 177,312 GitHub stars, last pushed on 6 September 2026, with a $14.5M Series A led by Nexus Venture Partners in August 2025.
Teams that want the option to self-host and can live without the cloud product's anti-bot layer.

Teams expecting the self-hosted build to match the cloud product's anti-bot capability. It does not.
Scraping sites that fight back on a tight budget: Stealth Mode and AI-schema /extract calls cost up to 5 credits against 1 for a basic scrape.
Tasks where something has to be clicked, typed or logged into rather than read.

What it actually does

Firecrawl is an API you point at a URL, and it hands back the page as clean text a model can read, with the ads, menus and markup taken out. That makes it the input stage for retrieval pipelines, agents and research tools: fetch the page, get usable text, embed it. Everything is metered in credits. The free tier is a recurring 1,000 credits a month, roughly 500 searches or 1,000 pages, with two concurrent requests and no card. Paid plans buy a bigger pool and more concurrency: Hobby is 5,000 credits and 5 concurrent requests at $16 a month billed yearly, Standard is 100,000 and 25, and Scale is 1,000,000 and 100. The credit arithmetic is where the bill surprises people. A basic scrape costs 1 credit, but Stealth Mode and AI-schema /extract calls cost up to 5, so a plan sized on easy pages drains five times faster once the hard ones start. The core is AGPL-3.0 with MIT SDKs and you can self-host it, but the self-hosted build does not include the managed anti-bot "fire engine" and stealth proxying the cloud has — which is the part that matters on sites that resist scrapers.

What the free tier actually buys

1,000 credits a month — roughly 500 searches or 1,000 pages — with two requests running at a time. It renews every month, unlike the one-time credits E2B and Browser Use give you once.

From the vendor's docs

Pricing and free-plan limits

Free tier: 1,000 credits a month — roughly 500 searches or 1,000 pages — with 2 concurrent requests and no card. It renews monthly.

Plan	Price	What you get
Free	Free	1,000 credits a month, about 500 searches or 1,000 pages No card required Limits: 2 concurrent requests A recurring monthly allowance, not a one-time signup credit.
Hobby	$16/month	5,000 credits 5 concurrent requests 100 requests a minute The rate billed yearly. We have not recorded a monthly-billing price.
Standard	$83/month	100,000 credits 25 concurrent requests 500 requests a minute The rate billed yearly.
Growth	$333/month	500,000 credits 50 concurrent requests The rate billed yearly.
Scale	$599/month	1,000,000 credits 100 concurrent requests The rate billed yearly.
Enterprise	Custom pricing	Quoted rather than listed No figure is published.

Prices checked on Sep 6, 2026 against the vendor's pricing page. Plans change often — confirm before you pay.

Three honest limitations

The self-hosted build is not the cloud product. A GitHub issue on the project records that it lacks the managed anti-bot "fire engine" and the stealth proxying the hosted version has, which is exactly what difficult sites need.
Credits are not all equal. A basic scrape costs 1, while Stealth Mode and AI-schema /extract calls cost up to 5, so costs escalate on hard sites.
The core is AGPL-3.0. Running a modified copy as a network service requires publishing your modifications. The SDKs are MIT.

Alternatives

Browser Use — Choose Browser Use instead if the job is to drive a browser — click, type, log in — rather than to read what a page says.
E2B — Choose E2B instead if what you need is a sandbox to run the code your model wrote, not a way to fetch pages.
Langfuse — Choose Langfuse instead if the pages already arrive fine and the open question is what your app did with them.

Getting started in 5 minutes

Start on the free tier: 1,000 credits a month, two concurrent requests, no card. That is enough to find out whether the pages you care about come back clean.
Scrape one page before you crawl a site. Credits are the unit of everything here, and a crawl spends them page by page.
Watch which calls you make. A basic scrape is 1 credit; Stealth Mode and AI-schema /extract calls cost up to 5.
Decide hosted or self-hosted early. Self-hosting is supported and the core is AGPL-3.0, but the managed anti-bot engine and stealth proxying stay in the cloud product.
Move up a plan when concurrency rather than credits is the wall: Hobby is 5 concurrent requests, Standard 25, Scale 100.

Firecrawl questions

Is Firecrawl free?: There is a genuine free tier, and it renews: 1,000 credits a month — roughly 500 searches or 1,000 pages — with two concurrent requests and no card. Paid plans start at $16 a month billed yearly for 5,000 credits and 5 concurrent requests, and run to $599 for a million credits and 100 concurrent requests. Enterprise is quoted.
What is a credit?: The unit everything here is metered in. A basic scrape costs 1. A 2026 review records that Stealth Mode and AI-schema /extract calls cost up to 5, which is why a plan sized on simple pages runs short once you start on sites that resist scraping.
Can I self-host Firecrawl?: Yes. The core is AGPL-3.0, the SDKs are MIT, and self-hosting is supported. What you do not get is parity: a GitHub issue on the project records that the self-hosted build lacks the managed anti-bot "fire engine" and stealth proxying the cloud product has. AGPL also means that running a modified copy as a network service requires publishing your modifications.
Firecrawl or Browser Use?: Read a page versus drive a browser. Firecrawl fetches a URL and returns clean text, with no browser session involved. Browser Use puts an agent in a real browser to click, type and log in. If all you need is what the page says, Firecrawl is the simpler and more predictable half; Browser Use earns its keep when the task needs a form filled or a login completed.
Is the project active?: 177,312 GitHub stars and a push on 6 September 2026. A $14.5M Series A led by Nexus Venture Partners closed in August 2025, and the company claims 350,000+ developers and 15x year-on-year growth — its own figures. Recent ships: a Developer Index over 70M code artifacts on 20 August 2026, and a Life Sciences Index over 41M papers on 13 August 2026.

Sources

Research-compiled, not hands-on tested. Compiled from official docs, the vendor's pricing page and community reviews as of Sep 6, 2026. We haven't tested it hands-on yet.