Log in and stay in
Reach gated sites and keep the session logged in across runs.
Browser MCP: give your agent a real browser
Claude and Cursor can already reason about the web. One config block gives your MCP client a cloud browser that reaches logins, CAPTCHAs, and scale, with a Live View replay of everything the agent did.
The Problem
MCP clients can think about the web, but not act on it
- Your MCP client can reason about a task, but it has no browser to actually carry it out.
- A local browser tool works on your machine, then breaks the moment the site puts up a login or a CAPTCHA.
- Bot detection blocks scripted requests within minutes, so the agent stalls mid-task.
- Sessions lose their login state between runs, so the agent starts from a logged-out page every time.
- Running more than one browser at once means standing up and babysitting your own infrastructure.
- When a run goes wrong, you have no record of what the agent actually saw or clicked.
The Solution
One config block, one cloud browser for your agent
- One config block: add the Browserbase MCP server to Claude, Cursor, or any MCP client and your agent has a browser. No harness to build, no infrastructure to run.
- Real cloud browsers: your agent navigates live pages the same way a person would, including the ones with no API.
- Verified: reach pages that turn away ordinary tooling, with managed CAPTCHA solving and no per-site tuning.
- Agent Identity: Web Bot Auth signs your agent’s requests, so sites can verify it instead of guessing.
- Persistent sessions: the agent stays logged in to gated sites across runs, so it picks up where it left off.
- Live View and replay: watch the session in real time and replay every action, so whatever your agent saw, you see.
- Parallel by default: run many browsers at once on Browserbase infrastructure, under one API key.
What your MCP client can do with a browser
Reach the whole web
Verified access and managed CAPTCHA solving, so the agent reaches pages ordinary tooling cannot.
Fill forms and submit
Navigate, click, type, and complete multi-step workflows on live pages.
See every step
Live View in real time plus a full replay of what the agent did.
Frequently Asked Questions
What is Browser MCP?
Browser MCP connects an MCP client, such as Claude or Cursor, to a real browser using the Model Context Protocol. The Browserbase MCP server gives your agent a cloud browser it can navigate, so it can act on the web instead of only reasoning about it.
How do I give my MCP client a browser?
Add the Browserbase MCP server to your MCP client with a single config block, then add your Browserbase API key. Cursor supports one-click install, and Claude Code adds it with one command. From there your agent can navigate, click, fill forms, and extract data.
Does it work with Claude and Cursor?
Yes. The Browserbase MCP server works with any MCP client, including Claude and Cursor. It runs as a hosted endpoint on Browserbase infrastructure, and you can also run it locally.
What happens when a site shows a login or a CAPTCHA?
Browserbase gives your agent Verified access with managed CAPTCHA solving, so it reaches pages that turn away ordinary tooling. Agent Identity signs your agent’s requests with Web Bot Auth, so sites can verify it. Persistent sessions keep the agent logged in across runs.
Can I see what the agent did?
Yes. Every session has a Live View you can watch in real time and a full replay of the agent’s actions, so whatever your agent saw, you see. Session logs and recordings make it straightforward to debug a run.
Can it run more than one browser at a time?
Yes. Browserbase runs browsers in parallel on its own infrastructure, under one API key, so your agent can work across many sites at once without you standing up and maintaining browser infrastructure yourself.