Use selenium-level browser control, from wherever you are and whichever device you're on. browselenium.online runs a real, human-like browser in the cloud and gives your code the same verbs you know from selenium — navigate, find, click, type, cookies, screenshots — over a thin API. The browser and all of its upkeep happen on our servers: a Raspberry Pi, a Lambda function or a coding agent all drive the same session.
"Add to your agent" copies a setup prompt your agent executes itself. "Get your API key" takes you to GitHub sign-in — free tier, no card.
browselenium exposes the selenium verbs you already know as MCP tools and routes every call to our web service, where a real, human-like browser does the actual work. No local browser, no driver binaries, no Xvfb, no headless flags, no RAM.
A Raspberry Pi, a serverless function, a cron job, a CI runner or an AI agent. If it can make an HTTPS call, it can run a browser session.
A real, human-like browser, tuned and maintained by us so it reads like an everyday visitor.
Visible text, DOM fragments, cookie jars and PNG screenshots come back to your process. Sessions persist across calls.
Because the browser lives on our infrastructure, serverless web scraping finally works: AWS Lambda, Google Cloud Functions and Cloudflare Workers can't run a real browser — but they can all call browselenium. The same goes for embedded boards and microservers where a full browser would not fit.
Point your agent at the hosted endpoint. On first use it authenticates with the API key from your dashboard and learns the toolset from the built-in skill.
{
"mcp": {
"browselenium-online": {
"type": "remote",
"url": "https://browselenium.online/mcp",
"enabled": true
}
}
}
That is the whole install: the browser lives on our infrastructure, so there is nothing to download, nothing to patch and nothing running on your machine — the tools are identical for every harness, from opencode to openclaw.
Thin client with no browser dependency: Raspberry Pi Zero, AWS Lambda, Cloudflare Workers, Docker, CI. Anywhere Python or MCP speaks, browsing works.
A real, human-like browser maintained by us — most Cloudflare interstitials just pass.
navigate, find, click, type_text, execute_js, tabs, cookies, screenshots — the workflow you know, minus the driver plumbing.
32 tools over remote HTTP or stdio. Works with opencode, openclaw, Claude Code, Codex, Cursor, Windsurf — anything that speaks the Model Context Protocol.
Type into forms, click through flows, export and restore cookies so your scraper or agent stays logged in across runs.
Viewport and full-page screenshots straight from the cloud browser — verify what the page actually looks like before you parse it.
Authenticate once and call get_skill — the agent loads the full playbook: workflows, limits, ground rules.
Per-tier request quotas and browser lifetimes, enforced server-side. No runaway scrapers, no surprise bills.
Custom user agents and locales per browser start. Self-hosted servers can also route sessions through their own proxies — hosted egress is service-controlled.
browselenium.online is one idea: use selenium-level browser control, from wherever you are and whichever device you're on. Most modern sites sit behind bot management — Cloudflare Turnstile, PerimeterX, DataDome, Akamai. Plain HTTP clients get a 403; vanilla headless browsers get flagged in milliseconds. browselenium gives you a remote, human-like browser tuned to look human, so web scraping, site monitoring, form automation and end-to-end checks just work — while your client code stays thin enough for serverless runtimes and embedded devices.
Typical uses: monitoring competitor prices from a Raspberry Pi, running scheduled scraping jobs in AWS Lambda, checking SERPs from Cloudflare Workers, letting a coding agent log into dashboards and read documentation behind JavaScript walls, and driving logged-in sessions that survive between runs via cookie export and restore.
Under the hood it is one small Python package speaking to one endpoint: an MCP server whose tools carry the selenium names you know. If you have used selenium before, you already know every verb — the only difference is that the browser runs on our infrastructure, maintained and upgraded by us, metered for fairness. Sign up on the dashboard, copy your API key, and start with 50 free requests a day; upgrade through Stripe when your scrapers or agents do.
Continue with GitHub on the dashboard — your API key is generated instantly (shown once, rotate anytime). The free tier needs no card: 50 requests a day. When you outgrow it, pick a tier in the dashboard and pay through Stripe; your limits rise the moment the payment webhook lands. Cancel anytime from the Stripe portal.
Yes. The client only sends HTTPS calls and receives JSON — it never launches a browser. A Raspberry Pi Zero, an AWS Lambda function, a Google Cloud Function or a Cloudflare Worker can all drive full cloud browser sessions. Memory and disk on your side stay flat.
With DIY automation you install a real browser, its driver and a pile of anti-bot fixes on every machine you deploy to — and you still get blocked. browselenium gives you the browser as a service: same selenium-style operations, none of the setup, and the browser tuning is maintained upstream for you.
Within a session the browser persists across every tool call, including cookies and history. You can also export the cookie jar with get_cookies and restore it later with set_cookies — even on a completely different machine or the next serverless invocation.
A hosted MCP endpoint with 32 browser tools: navigate, read text and HTML, find and click elements, type, press keys, scroll, run JavaScript, manage tabs, save screenshots and cookies. The built-in get_skill tool teaches the agent the full workflow on first contact. Works with opencode, openclaw, Claude Code, Codex, Cursor, Windsurf and any other MCP-capable client.
Use it the way a human would use a browser: respect site terms, robots.txt and applicable law, and keep request rates reasonable. The service enforces per-tier quotas and browser lifetimes to encourage fair use. See the terms of service for details.
Every tier ships the full toolset. Limits are requests per UTC day and maximum browser lifetime per session. Create an account, copy your API key from the dashboard, upgrade through Stripe — higher limits apply the moment the webhook lands.