FAQ
How Browse Bridge works, what the agent can do, whose IP a site sees, whether agent clicks are trusted events, what happens on a sensitive site, that the tab stays in this Chrome profile, how companies install it, what the takeover codes mean, and what stays your responsibility. Add to Chrome
How Browse Bridge works
- Install the Chrome extension and sign in to Browse Bridge in the browser you already use.
- Open a site yourself. Complete login, MFA, CAPTCHA, and any other human checks. Browse Bridge never solves or bypasses those.
- Share only the tab you want an agent to use. That tab stays in this Chrome profile and keeps your usual home or work IP. An unshared tab stays private.
- Connect an AI agent with a Browse Bridge API key from Connections. The agent talks to Browse Bridge servers, not to the website, your cookies, or your password manager.
- Give the agent control when you are ready. It can then read the page, click, type, navigate, and request an on-demand screenshot of that tab.
- Take control at any time from the extension popup, side panel, toolbar badge, or Ctrl+Shift+Period. Queued commands are cancelled. Cross-origin navigation waits for your permission.
Questions
What are agent actions?
An action is something the agent does to the shared tab: click, type, clear, select, check, uncheck, focus, press a key, scroll, navigate, go back or forward, reload, or wait. Each of those counts toward your monthly action cap. Reading the page structure and asking for status do not. Screenshots have a separate monthly cap. You see the running totals on Overview and Billing. The History page lists a short summary of what the agent requested, without storing typed secrets.
What are screenshots, and how are they used?
Screenshots are for the agent, not a gallery for you. When the structured page view is not enough — a chart, a canvas, or a custom control — the agent can request one on-demand picture of the visible tab. Browse Bridge is not a screen recorder. The image is stored only so the connected agent can look at that frame, then it is deleted. Free keeps screenshots for 24 hours. Pro keeps them for 7 days. Overview and Billing show how many you have used this month, not the pictures themselves.
What happens to History on each plan?
History is a log of agent activity without typed secrets. On Free, those events are kept for 24 hours. On Pro, they are kept for 90 days. Screenshot files expire on their own schedule, even if a History line about the request is still visible. Team and Business are coming soon and are not sold yet. After the retention window, the records are deleted.
Does Browse Bridge let me ignore a website’s terms of service?
No. Browse Bridge is a tool for sharing a tab you already control. It does not grant rights to scrape, automate, or access any site. You must follow the terms of the websites you open, the AI service you connect, and the law where you use the product.
Who is responsible if the agent does something I am not allowed to do?
You are. You choose the account, the tab, the agent, and when to give it control. Signup and Enable each require an explicit confirmation that you will follow site terms. If a site forbids automated use, that limit still applies when you hand the tab to an agent. Browse Bridge does not review or approve your destinations.
Will Browse Bridge solve CAPTCHAs or bypass bot detection for me?
No. Login, MFA, CAPTCHA, Cloudflare Turnstile, and similar checks stay in your hands. If a challenge appears, the agent pauses with CHALLENGE_DETECTED and does not retry through it. You complete the check, then Enable the agent again.
Does Browse Bridge mark the page so a site can detect the extension?
The agent script runs only while a tab is shared, in an isolated world the page cannot see. It does not stamp data attributes or enumerable globals onto the document. Disable tears the script down and clears in-memory element IDs. Clicks and typing do not attach Chrome's debugger. A screenshot may attach it only long enough to capture a tab Chrome's visible-tab API cannot see, then detaches. browser_handle_dialog may attach it only long enough to dismiss a JavaScript dialog that is already open, then detaches. Sites can still notice automated clicking the same way they notice a person using a mouse.
Does Browse Bridge click as fast as the agent sends commands?
No. Click, key, navigation, and screenshot requests are paced on the shared tab. If an agent sends too many at once, Browse Bridge returns RATE_LIMITED with a wait time instead of hammering the page. A CAPTCHA or bot challenge is not retried. Browse Bridge does not fake mouse movement or spoof WebDriver.
Does an agent click count as a real user click?
Clicks use the page's DOM click() method and typing sets the field value. Those events are not isTrusted. File pickers, some clipboard and payment prompts, and Chrome permission requests still need a click from you. Browse Bridge does not spoof isTrusted or attach the debugger to fake a mouse.
What happens on a bank or login site?
Sensitive-site mode asks you to Approve each click, type, or navigation. The popup shows Waiting for Approve, which is not a CAPTCHA or CHALLENGE_DETECTED pause. Approve runs that action; Deny cancels it. You can turn the mode on from the extension, including for local test pages. Keep banks and logins in this Chrome profile — do not open a second automation profile for them. Browse Bridge still will not solve challenges or spoof trusted clicks.
Whose IP address does a website see when an agent uses my tab?
The website sees this Chrome profile's network — typically your home or office IP — because the default path is the tab you already have open. Agent control uses Browse Bridge's MCP host (mcp.browsebridge.com), a datacenter path between the AI and Browse Bridge only. That host does not load the page and must not appear in the site's DevTools Network log. Browse Bridge does not spoof, rotate, or proxy page IPs. If you run this browser in a datacenter or through your own VPN, sites will see that address instead.
How do I connect Cursor or another MCP client?
Open Connections in the dashboard. Generate an API key — the default name is API Key — then Copy the MCP URL (https://mcp.browsebridge.com/mcp) and the full bbk_ key into the client. The full secret is shown once. Cursor and other MCP clients send it as a Bearer token. You can rename a key later; that name is stored and still shown after you reload. Revoke a leaked or unused key so that Bearer token stops working immediately.
Where are the Terms of Service and how do confirmations work?
The Privacy Policy, Terms of Service, Acceptable Use Policy, and frozen prior versions are public. Each frozen id has its full text at /legal/versions. Signup and Enable show a versioned confirmation that must be Agreed before an account or shared tab is created. Disagree cancels signup or leaves the tab unshared. Browse Bridge does not solve CAPTCHAs, read cookies for agents, or run agent JavaScript. Report abuse or get support from the public form linked in the footer.
Can I share a tab that is signed in as someone else?
Only share sessions you are authorized to use. Handing another person’s logged-in tab to an agent does not make that use allowed. You remain responsible for the accounts and data the agent can reach.
Can my company force-install Browse Bridge, and does it collect SSO passwords?
Production installs come from the Chrome Web Store, not an unpacked developer folder. IT can allowlist or force-install the Browse Bridge extension with MDM. If a proxy blocks unknown hosts, allow Browse Bridge's own API hosts only — not the websites in the tab. Sign into those sites, including SSO, in your own Chrome. Browse Bridge does not scrape passwords, SSO cookies, or SAML assertions.
How does the dashboard know if the Chrome extension is installed?
The dashboard asks this Chrome profile for the Browse Bridge extension. That is not the same as a device showing online on Devices — online is a signed-in heartbeat from any of your installs. If this browser has no extension, the dashboard links to the Chrome Web Store. The ping only checks that the extension is present; it does not read controlMode, cookies, or tab contents.
Should I use a separate Chrome profile so the agent looks less like me?
No. The shared tab stays in this Chrome profile — the same cookies, IP, and login you already use. controlMode is not exposed on the page window. For banks and other sensitive sites, stay here and use Sensitive site instead of a second automation profile, headless browser, or extra user-data-dir.
What should I do when the agent stops and shows a code?
The popup and MCP use the same four codes. CHALLENGE_DETECTED: complete the check yourself, then Enable and Agent — Browse Bridge will not retry through it. ORIGIN_NOT_ALLOWED: Allow the origin in the extension, or Disable and Enable if capture permission is missing. CONTENT_SCRIPT_UNAVAILABLE: Enable the tab again; Browse Bridge reattaches the agent. HUMAN_HAS_CONTROL: click Agent when the agent should continue. Waiting for Approve is a separate sensitive-site wait, not a challenge.