Free
Test every feature.
Full platform to try, no card needed
- Every feature unlocked
- API · CLI · SDKs · MCP · Agent Toolkit
- All output formats
- Community support
Start free. Every plan unlocks the full API, all output formats, and the whole toolkit. Pay for the volume you run.
Start free. Pick the volume your team needs. Every plan includes the full platform, one subscription.
Test every feature.
Full platform to try, no card needed
Launch fast, scale without setup.
250K requests/month · 20 concurrency
Reliable scraping for real workloads.
1M requests/month · 50 concurrency
Scale reliable scraping across a team.
3M requests/month · 100 concurrency
Committed volume and support.
Unlimited volume, tailored to you
Volumes are standard requests. Protected pages and browser rendering use more, see the FAQ. Prices in USD.
Set your monthly volume across the platform. We point you to the plan that covers it.
Eight primitives
Each primitive is a verb you call from the API, CLI, MCP, or any SDK. Same surface, same Recovery Engine, one subscription.
Eight building blocks your team can mix. Same platform, same recovery layer, one subscription for every workflow.
fetch(url)
One call · any page
Clean content from any URL. Standard, dynamic, or protected pages, HTML, Markdown, JSON, screenshot.
Clean content from any page. Standard, dynamic, or protected sources, delivered as readable text, structured data, or a screenshot.
extract(html, schema)
Page in · fields out
Schema-validated JSON from messy, dynamic, and protected pages. No brittle parsers.
Structured data in the format your team specifies, pulled from messy, dynamic, and protected pages.
crawl(seed)
Start point · full map
Discover URLs, pagination, listings. Walks the site graph so your agents and pipelines know where the data lives.
Find every page that matters across a site. Listings, pagination, sub-sections, so your team always knows where the data lives.
batch.submit(jobs)
Many jobs · one run
Run large async web data jobs reliably. Queues, retries, webhooks, observability.
Run large workloads overnight or on schedule. Automatic retries, progress visibility, alerts when results are ready.
sessions.open()
Real browser · on demand
Operate dynamic websites with a real browser when pages need clicks, forms, login, or session state.
Handle live websites that need logins, clicks, or forms. The platform stays logged in across long jobs.
monitor(url, every)
Pick a page · watch it
Track pages, listings, datasets, and changes over time so your systems always work with fresh data.
Track pages, listings, and changes over time. Your team gets an alert when something actually moves.
toolkit.skill()
Drop-in · works anywhere
CLI, MCP server, SDKs, skills, templates, recipes. Every primitive, agent-ready.
Templates and connectors so your team's AI assistant can use every primitive out of the box.
kraken.dataset()
Describe it · receive it
No-code dataset builder. Describe what you need, get the structured data, no engineering required.
No-code dataset builder. Describe what your team needs, receive clean data, no engineering tickets.
Why Zenrows
Zenrows is built for the web that fights back: protected sites, browser workflows, recurring jobs, high-volume pipelines, and the AI surfaces that consume the data.
Zenrows is built for the sources that fight back: protected sites, login flows, scheduled refreshes, high-volume runs, and the AI assistants your team already uses.
You only pay for successful requests. Failed attempts aren't charged, and 404 or 410 responses count as usable results.
Your team only pays for successful requests. Failed attempts aren't charged, and 404 or 410 responses count as usable results.
Hard protected pages, browser workflows, residential network. The infrastructure stack lives underneath the verbs.
Hard-to-reach sources, login workflows, regional access. The complex parts live under the surface so your team doesn't manage them.
Eight primitives, one subscription, the same Recovery Engine behind every call. No glue code per surface.
Eight primitives, one balance, the same recovery layer behind every job. No separate tools, no separate budgets per use case.
MCP server, CLI, SDKs. Every primitive ships agent-shaped, with auth, logs, and observability.
Connectors and templates so your team's AI assistant plugs into every primitive out of the box, with permissions and activity logs included.
FAQ
Pricing depends on the complexity of each request, so you only pay for the tech you actually need. A basic request is the standard rate per 1,000. JS rendering costs 5x, premium proxies 10x, and both together 25x. On the Business plan that works out to $0.10 per 1,000 basic requests, $0.50 with JS rendering, $1.00 with premium proxies, and $2.50 for both.
Never. You only pay for successful requests. Not Found (404) and Gone (410) responses count as successful results.
Yes. A longer billing period lowers the monthly price automatically: quarterly saves 5%, semi-annually 7.5%, and annually 10%. On the Business 300 plan that is $285, $277, and $270 per month.
We warn you before you reach it. If you do hit the limit, scraping pauses and you can resume instantly by adding extra usage. Extra usage is prepaid, worth 15% of your plan, and you can add it up to 4 times a month. On Business 300 that is up to four top-ups of $44.99 each, with no lost data or configuration.
Yes. Upgrades take effect immediately, with billing adjusted accordingly. Downgrades take effect at the end of your current billing cycle. No penalties, no hidden fees.
Concurrency is how many requests run at the same time. The limit scales with your plan, from 20 on Developer to 50 on Startup and up to 400 on the largest Business tier.
Same primitives, same API. Set the protection level and Zenrows routes through the right stack (residential network, fingerprinting, retries, the Recovery Engine) so your code doesn't change between an open page and a hard-protected one.
Same toolkit, same workflow. Set the protection level and Zenrows picks the right route in the background. Your team's workflow looks the same whether the source is open or hard-protected.
Yes. The Agent Toolkit ships an MCP server, a CLI, and SDKs that snap into Claude Code, Cursor, Codex, Windsurf, Gemini CLI, and any MCP-compatible client. Every primitive is exposed as a tool, so agents call them natively with auth + observability built in.
Yes. The Agent Toolkit plugs into Claude, Cursor, Codex, Windsurf, Gemini, and other AI assistants your team uses. Every primitive is a ready-made tool, with permissions and activity logs included.
Yes, on Enterprise. SLA on uptime and success rate, dedicated capacity with committed usage, SSO/SAML, audit logs, DPA, security review, a named account manager, a solutions engineer, and a quarterly architecture review.
Ready when you are
Test every feature for free. Upgrade only when you cross the volume, the protected pages, or the browser workflows that need it.
Test everything for free. Upgrade only when the data you need crosses the volume, the protected sources, or the workflows that justify it.
Start free. No credit card required.