AnyCrawler
AnyCrawler is web access infrastructure for AI agents and applications: a single production API that searches the live web, fetches or renders pages into clean Markdown, captures screenshots, and meters every operation through flexible credits.
What is AnyCrawler?
AnyCrawler presents itself as web access infrastructure rather than a scraping API, and the distinction is the product: discovery, extraction and visual evidence sit behind one authenticated interface, so an AI system never has to run its own crawlers, browsers and parsers.
Three API families rest on that base. Search is the discovery layer, a POST to /v1/search across five channels (page, images, news, videos and scholar), billed at 20 credits per ten results, used to build a shortlist when the target URL is not yet known. Page crawl is the extraction layer: one endpoint, POST /v1/crawl/page, whose method field switches between two very different routes. Fetch takes the lightweight HTTP path with no browser start-up, at a base cost of 2 credits, and suits documentation, articles and reference pages. Render opens the URL in a browser, waits for client-loaded content and extracts the rendered DOM for 10 credits, which is what single-page apps, client-side routes and hydrated pricing tables require. Screenshot is the evidence layer, a POST to /v1/crawl/screenshot that stores a PNG and returns its URL.
Whichever route you take, the response shape holds: ok, requested_url, canonical_url, final_url, status_code, credits_used, and a results object carrying title, description, markdown, markdown_tokens and, on request, metadata, links and media. The vendor's headline claim for the cleaning step is blunt: HTML at 100 percent, Markdown at under 10.
Access runs on API keys sent as a bearer token or an x-api-key header, with console sign-in through Google or GitHub. A TypeScript SDK ships as @anycrawler/sdk, while an agent skill and the API documentation live in public GitHub repositories. Error codes are enumerated for production handling: 400, 401, 402 for exhausted credits, 403 and 429. Plans differ less in features than in throughput, from 30 to 1,200 fetch requests per minute and 1 to 20 concurrent browsers, and a free endpoint lets anyone try a crawl before signing up.
What it does
- Search the live web across five channels: pages, images, news, videos and scholarly sources
- Fetch a known URL over a lightweight HTTP path and receive normalized Markdown for 2 credits
- Render JavaScript-heavy pages in a real browser when a plain HTTP response is not enough
- Strip navigation, scripts and layout noise so a model receives usable context instead of raw HTML
- Capture a stored PNG screenshot of any public page for audits and human review
- Return traceable usage fields on every call: status code, final URL, credits used and request identifier
- Test the crawl endpoint with no account at all through the free, rate-limited URL
When to use AnyCrawler / When not to
A quick filter to help you decide if AnyCrawler is the right fit.
When to use AnyCrawler
- Engineering teams shipping AI agents that need live web context without building and maintaining their own crawlers, browsers and parsers
- Developers assembling RAG pipelines who ingest documentation, changelogs, help centers and API references at a low per-page cost
- Research and monitoring products that track pricing, policy, product or competitor pages on a repeatable schedule
- Indie developers and early-stage founders validating an idea on the free tier before committing to a 5 USD monthly plan
- Academic and technical researchers who need a scholar channel for papers, citations and references alongside general web search
When not to use AnyCrawler
- Non-technical users hoping for a point-and-click scraper: every call is an authenticated HTTP POST, and the console only handles keys, usage and logs
- Teams that need a documented GDPR posture, since the site never mentions the regulation and publishes no DPA, no subprocessor list and no Article 27 representative
- Organizations handling regulated or highly sensitive data, which the privacy policy explicitly says the product is not marketed for
- Anyone looking to bypass paywalls, CAPTCHAs, logins or access controls, all of which the terms of service prohibit outright
- Buyers who need a contractual uptime commitment, as the terms disclaim any promise of uninterrupted service and no SLA is published
How to use AnyCrawler
A typical end-to-end flow, from setup to results.
- Try a crawl with no account at all: send a URL to the free endpoint at api.anycrawler.com/free/v1/crawl, or use the form on the Free Crawl page
- Create a workspace by signing in with Google or GitHub, which credits the account with 5,000 one-time credits
- Collect your API key from the console and send it as an Authorization bearer token or an x-api-key header
- Pick the path before the call: search when the URL is unknown, fetch when the HTML already carries the content, render when JavaScript builds the page, screenshot when a human needs to see it
- Call POST /v1/search with the channel you need to assemble a shortlist of candidate sources
- Call POST /v1/crawl/page with the URL and method=fetch, adding include_metadata, include_links or markdown_variant as required
- Switch the same call to method=render for JavaScript pages, tuning browser_wait_until and accept_cache
- Call POST /v1/crawl/screenshot with full_page when the workflow needs a stored image, then keep the returned URL with your audit record
- Read credits_used, status_code and the request identifier on every response, both to debug and to keep monthly consumption under control
- Move to production through the @anycrawler/sdk TypeScript client or by installing the agent skill from GitHub, upgrading the plan as throughput demands
Pros & Cons
Pros
- One request shape covers four jobs: only the method field separates a cheap fetch from a full browser render
- Billing follows the work performed, so static pages are never charged at browser rates, 2 credits against 10
- The barrier to entry is close to zero: a permanent free tier, 5,000 credits on signup, a no-account test endpoint and a first paid plan at 5 USD a month
- Unit cost falls sharply with volume, from 0.000333 USD per credit on Agent Lite to 0.0000667 USD on Scale
- Responses are traceable by design: status code, final URL, canonical URL, credits used, token counts and a request identifier
- Error codes are explicit enough to drive retry logic in production, including a dedicated 402 for exhausted credits
- A scholar search channel for papers and citations is unusual among general-purpose crawling APIs
Cons
- The publisher is anonymous: no legal entity, no postal address and no legal notice, about or contact page anywhere on the site
- Data governance is undocumented, with no GDPR mention, no DPA, no subprocessor list, no Article 27 representative and no security certification
- The privacy policy states outright that no jurisdiction-specific transfer framework and no country-by-country operations list are published, leaving hosting unknown
- Retention is described in kind but never in figures: no window is given for request logs, cached responses or stored screenshots
- There is no availability commitment at all, the terms disclaiming uninterrupted service, and no SLA is published
- The domain was registered in March 2025 and the Wayback Machine holds no capture, so there is no verifiable track record behind the product
- Support runs through a single generic email address, with no ticketing, no chat and no documented response time
Pricing & Plans
AnyCrawler offers a permanent free plan, listed at 0 USD per month and granting 5,000 one-time credits. The cheapest paid entry point is the Agent Lite plan, at 5.00 USD per month. No time-limited free trial is advertised; instead, a 30-day money-back window generally applies to a first self-serve purchase, handled by Paddle as merchant of record.
- 5
- 000 one-time credits
- 30 fetch requests per minute
- 1 concurrent browser
- presented as a plan for early validation
- 15
- 000 credits per month at 0.000333 USD per credit
- 60 fetch requests per minute
- 2 concurrent browsers
- with persistent top-ups sold separately
- 80
- 000 credits per month at 0.000250 USD per credit
- 180 fetch requests per minute
- 3 concurrent browsers
- 960
- 000 credits per month at 0.0000833 USD per credit
- 480 fetch requests per minute
- 8 concurrent browsers
- 3
- 000
- 000 credits per month at 0.0000667 USD per credit
- 1
- 200 fetch requests per minute
- 20 concurrent browsers
- a fetch request costs 2 credits
- a render request 10 credits
- and a search returns ten results for 20 credits
Data, GDPR & hosting
A consolidated view of how AnyCrawler handles your data.
GDPR overview
There is no GDPR implementation to report, because the site never mentions the regulation. Across the ten pages collected, the full privacy policy and terms included, the words GDPR, data processing agreement and subprocessor do not appear once, neither in the visible text nor in the archived HTML. No Article 27 EU representative is designated, no data protection officer is named, and no security certification such as SOC 2 or ISO 27001 is claimed. Section 9 of the privacy policy does gesture at data subject rights, offering access, correction, deletion, restriction and objection "depending on where you live", subject to identity verification and lawful refusal; formal requests go to the generic support mailbox. Section 8 states outright that no jurisdiction-specific transfer framework and no country-by-country operations list are published. Treat this as undocumented rather than compliant.
Who owns the data?
The terms draw a clear line. You keep responsibility for the prompts, payloads, URLs, settings and outputs you send through the service, and AnyCrawler explicitly does not claim ownership of third-party web content merely because it processed that content on your instruction. In exchange, you grant the vendor the limited rights it needs to host, transmit, cache, process and display your submissions and their results, in order to operate, secure and improve the service. AnyCrawler owns its own software, branding and documentation, and grants you only a limited, non-exclusive, non-transferable right of use. The privacy policy states that personal information is not sold, while allowing sharing with service providers, disclosure under legal process, and transfer in a merger or asset sale.
Reuse rights
Nothing in the terms grants you a license over the third-party content you retrieve. AnyCrawler hands the question straight back to you: you remain responsible for the targets you choose, the instructions you send, the outputs you store or redistribute, and the way those outputs feed your business or your AI systems. You are equally expected to assess the target site's own terms, its robots behavior, contractual restrictions, privacy obligations, export controls and any sector-specific regulation. The vendor makes no representation that a given site may lawfully or contractually be crawled, rendered, copied, cached, searched or transformed, and states plainly that a page being technically reachable does not mean your intended use is permitted. Reuse, in short, is yours to justify rather than the vendor's to authorize.
Data retention & training
Hosting summary
AnyCrawler publishes no hosting information, and says so in as many words. Section 8 of the privacy policy states that the site does not currently publish a jurisdiction-specific data transfer framework or a country-by-country operations list, and reserves the right to add such detail later if its operational footprint or legal obligations require it. No hosting country, region or cloud provider is named anywhere on the site. Security is described in general terms only: administrative, technical and organizational measures, authentication controls, network protections, access restriction, logging and operational review, with no certification cited and an explicit refusal to promise absolute security. Two network facts deserve to be kept apart from these claims, because they say less than they appear to: the domain resolves to a Cloudflare anycast address, which identifies an edge node rather than a storage location, and the registrar is Alibaba Cloud's HiChina, which describes the registration channel rather than where anything runs. Anyone with a data residency requirement should treat hosting as undisclosed and ask directly.
Things to keep in mind
Risks and trade-offs to weigh before adopting AnyCrawler.
- You cannot identify who you are contracting with: no legal entity, no address and no named representative appear anywhere, leaving you nowhere to turn in a dispute
- The terms name no governing law and no competent court, while the domain is registered through a Chinese registrar, so the applicable jurisdiction is genuinely unknown
- The entire compliance burden of crawling sits on you: target site terms, robots behavior, copyright and personal data are all explicitly your problem
- Target URLs, search queries, extracted Markdown, screenshots and diagnostic logs may be cached, logged or retained, and no retention period is ever quantified
- Nothing on the site says whether customer data feeds model training, and no opt-out is documented, because the subject is simply never addressed
- An unbounded agent loop can burn a monthly credit allowance quickly, and no documented spending guardrail stands in the way
- Building a production dependency on a service with no SLA, no uptime promise and barely a year of domain history carries real continuity risk
Setup & Integrations
Technical difficulty
Low for a developer, impassable for anyone else. A first crawl needs no setup at all: the free endpoint accepts a URL from a browser form or a plain GET. Beyond that, everything is an authenticated HTTP POST with a JSON body, so the real requirement is comfort with API calls rather than any installation. Copy-paste curl and TypeScript examples appear on each endpoint page, a published SDK shortens the work further, and an installable agent skill covers the agent case. There is no graphical crawling interface; the console only manages keys, usage and logs.
Deployment
Integrations
Behind AnyCrawler
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
What exactly is AnyCrawler?
Is it just another web scraping API?
How does the credit billing work?
When should I use fetch, render, search or screenshot?
Can I try it without creating an account?
Is there a free plan, and what does the cheapest paid plan cost?
How do I authenticate, and is there an SDK?
Where is my data hosted?
Who is responsible for the legality of what I crawl?
Who processes payments, and can I get a refund?
Should you pick AnyCrawler?
AnyCrawler is an unusually legible product wrapped in an unusually opaque company.
On the technical side, everything a buyer needs is on the page. Endpoints are named, credit costs are stated per operation, request options are listed, response fields are enumerated and error codes are documented well enough to drive retry logic. The economics are just as transparent: a fetch costs 2 credits where a render costs 10, and the unit price falls fivefold between the entry plan and the top one. The barrier to trying it is close to zero, with a free endpoint that needs no account, 5,000 credits on signup and a first paid plan at 5 USD a month. A scholar search channel, rare among general-purpose crawling APIs, is a genuine differentiator for research workflows.
On the corporate side, almost nothing is on the page. There is no legal entity, no postal address, no legal notice and no contact route beyond one support mailbox. The GDPR is never mentioned, no DPA or subprocessor list exists, no security certification is claimed, and the privacy policy states outright that hosting jurisdictions are not published. Retention is described in kind but never in figures, and the terms promise no availability whatsoever. The domain dates from March 2025 with no Wayback capture behind it, so there is no track record to lean on.
The verdict follows that split. For a developer wiring web access into an agent, the price, the documentation and the free entry make it easy to evaluate on its merits, and paying through Paddle limits the exposure. For anyone with regulatory obligations, a procurement checklist or a production dependency that needs a named counterparty, the missing half of this site is disqualifying until it is filled in.
- Choosing a selection results in a full page refresh.
- Opens in a new window.