Browserless MCP Server
Browserless.io MCP server: a cloud browser for the agent with page reads, site crawls, Lighthouse audits and click-driven sessions
- 18 ★ on the repository
- Last commit
Broad access
Why this level
- The agent drives a real browser: it clicks, types and downloads files, and saved sign-in profiles give it access to your accounts on sites
- Captchas, proxies and operating system emulation can get around site bot protection, which may break site terms or the law
- Pages and profile data pass through a third-party cloud service, and the token opens your plan and spends units
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/browserless-mcp-serverDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The agent drives a real browser: it clicks, types and downloads files, and saved sign-in profiles give it access to your accounts on sites
- Captchas, proxies and operating system emulation can get around site bot protection, which may break site terms or the law
- Pages and profile data pass through a third-party cloud service, and the token opens your plan and spends units
This is third-party code. Review the repository files before installing.
What it does
The hosted version runs at `https://mcp.browserless.io/mcp`, so nothing needs installing. The agent gets a browser in the cloud and reads pages through it. The `browserless_smartscraper` tool reads one page and returns Markdown or HTML, coping with heavy JavaScript. `browserless_crawl` walks a site and collects every page it finds, `browserless_map` builds a URL list from the sitemap, and `browserless_search` searches the web and can scrape the results. For interface work there is `browserless_agent`: a persistent session in which the agent snapshots the page, plans steps and runs clicks, typing and scrolling in batches. It supports multiple tabs, screenshots, file upload and download, captcha solving in the cloud version, and routing through datacenter or residential proxies. `browserless_function` runs your Puppeteer code, `browserless_performance` runs a Lighthouse audit, and `browserless_export` returns a page in its native format, for example PDF. Support tools show the plan and unit usage, request logs, recorded sessions and saved sign-in profiles. Built-in skills teach the agent shadow DOM, cookie consent dialogs, modals and dynamic content.
Who it is for. Developers and analysts who need web access for an agent in Claude Code, Cursor or VS Code: reading heavy sites, crawls, page speed checks and click-driven scripts.
Good fit when
- A page is built with JavaScript and a plain fetch returns an empty shell: `browserless_smartscraper` returns ready Markdown
- A whole site has to be walked: `browserless_map` shows the URLs and `browserless_crawl` collects pages with depth and path limits
- Before a release you want a page speed estimate: the Lighthouse audit runs with phone or desktop emulation and performance budgets
- A flow behind a login is needed: the agent works in a saved profile and hands control to you for verification
Not a fit when
- The page can be fetched with a plain request: a cloud browser spends Browserless plan units
- You are not ready to hand page data and sign-in profiles to a third-party cloud and only a local browser fits
- You need offline work: the hosted server requires internet access and a Browserless token
Example request
Read the pricing page with Browserless, return it as Markdown and compare it with the week-old cacheMCP
- Transport
- http
- Authentication
- API key
Environment variables
BROWSERLESS_TOKENrequired, secret- Browserless.io token, required for self-hosting
BROWSERLESS_API_URL- Address of your own Browserless instance, defaults to `https://production-sfo.browserless.io`
BROWSERLESS_TIMEOUT- Request timeout in milliseconds, defaults to 30000
Limitations
- A Browserless.io token is required, and a free tier exists. Every call spends plan units and the proxy is billed by traffic.
- Captcha solving works in the cloud version only. State and city geotargeting needs paid plans and otherwise returns 401.
- The token goes in the `Authorization: Bearer` header or the `?token=` parameter in the address. An address with a token must not be published.
- The server license is SSPL-1.0, and self-hosting builds a Docker image with `BROWSERLESS_TOKEN`.
How to disable. Remove the `browserless` block from the client MCP configuration and restart it. The token can be replaced or revoked in the Browserless.io dashboard.
FAQ
How do I connect Browserless to Claude Desktop or Cursor?
Add a `browserless` block with the address `https://mcp.browserless.io/mcp` and your token to the MCP configuration: either the `Authorization: Bearer` header or the `?token=` parameter. No local install is needed.
Which tools does the server have?
They include `browserless_smartscraper`, `browserless_search`, `browserless_map`, `browserless_crawl`, `browserless_performance`, `browserless_function`, `browserless_export` and `browserless_agent`. There are also tools for skills, profiles, plan, usage, sessions and logs.
Can I run the server myself?
Yes. Build the image with `docker build -f docker/Dockerfile -t browserless-mcp .` and run it with `BROWSERLESS_TOKEN` and, if needed, `BROWSERLESS_API_URL` for your own Browserless instance. The client connects to `http://localhost:8080/mcp`.
What does `browserless_agent` do?
It drives a persistent browser session: snapshot, plan, batch of actions and a new snapshot. It handles multiple tabs, screenshots, file upload and proxies. The `proxy` parameter sets datacenter or residential addresses, and `emulationOs` picks the browser operating system.
How do I hand the browser to a person?
A handoff flow covers it: `browserless_live_url` with the `handoff` action opens the existing session in an interactive viewer. Agent commands are blocked while a person is in control. Work resumes after the `userDone` reason.
Discussion
No questions or reviews yet.
Related
MCP serversHyperbrowser MCP
Hyperbrowser MCP Server
The official MCP server for Hyperbrowser cloud browsers: page scraping and crawling, structured data, Bing search and three browser agents
An MCP server that opens pages in headless Chromium through Playwright and hands the agent the main text as Markdown, even from JavaScript-heavy sites
A separate browser for agents with a skill and MCP: Claude Code and Codex work in your logged-in accounts, and every session is recorded for replay
An MCP server and Chrome extension that let the agent drive your already open browser with your logged-in sessions, all locally
An MCP server that lets an agent drive a real browser: open pages, click and fill forms through Playwright
MCP gives an agent a Playwright browser where a person can take over the same session and save an authentication profile
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Browserless MCP Server (this card) | yes | API key | 18 | |
| Browserbase Skills | no | 3,754 | skills for a different browser cloud, Browserbase, rather than the Browserless MCP server | |
| Hyperbrowser MCP | yes | API key | 793 | an MCP server for Hyperbrowser cloud browsers with three ready browser agents |
README in short
The page opens with a quick start on the hosted server and a tool table. It then covers session replay, the Lighthouse scorecard, handoff to a person, built-in skills, proxies and the OS persona. Below come client configurations, self-hosting, environment variables and diagnostics.