Manus Browser: How to Take Over When a Task Gets Stuck

How-to — 2026-10-25 — by Mahmoud Zalt

What to do when Manus hits a login, CAPTCHA or verification step: take over the browser, save the session, or run the task in your own browser instead.

Every autonomous browsing agent runs into the same wall eventually. The task is going fine, then it reaches a sign-in page, a cookie wall or a CAPTCHA, and the run just sits there. This is not a bug in Manus so much as the shape of the open web, and there is a documented way through it.

This guide covers what the Manus cloud browser is, how to take control of it mid-task, how saved logins work and where to delete them, when to run the task in your own browser instead, and the blocks that no amount of clever prompting will get past.

What is the Manus cloud browser?

A real browser that Manus drives on its own machine. Its documentation describes the cloud browser as visiting websites, clicking buttons, filling forms, extracting data and completing multi-step workflows, which is what separates an agent from a text-only assistant.

It can also work while signed in. Manus states you can log into personal or work accounts inside that browser so it can carry out authenticated tasks such as checking email or updating a CRM record. That capability is exactly why the login moments matter so much.

How do you take over the browser mid-task?

Open the browser view from the running task and drive it yourself. Manus's help centre on browser access issues says the control sits in the upper right corner of the task, and that clicking it lets you manually take over the browser and sign in to the site you need so the CAPTCHA stops blocking the run.

In the cloud browser documentation, the same mechanism is described from the other side: when Manus meets a complex verification, meaning an SMS code, a CAPTCHA or a multi-factor prompt, it triggers a take over request. You complete the verification and hand control back, and it carries on with the task from where it stopped.

Getting a stuck Manus task moving again

Watch for the stall rather than the failure — A blocked run often does not error, it just stops progressing. Keep the run panel open on tasks that involve signed-in sites so you notice the pause in the first minute rather than the fortieth.
Open the browser control in the task — Manus's help centre describes finding the browser option in the upper right corner of the task and clicking it to take manual control.
Complete the step yourself — Sign in, clear the CAPTCHA, enter the code. Manus's cloud browser documentation names SMS codes, CAPTCHAs and multi-factor authentication as the verification types that prompt a take over.
Hand control back — The documented pattern is that you temporarily control the browser and then return control so Manus continues the same task rather than restarting it.
Decide whether to save the session — Manus's help centre states it will ask you to confirm whether to save the login status for that website, so it can reach the site independently next time without you assisting.
If the site keeps blocking, switch browsers — Manus documents an alternative that runs in your own browser using your residential IP address, which it recommends for security-sensitive sites and for services that block data centre traffic.

The habit worth building is the first step. An agent that is stuck is still an agent that is consuming your attention later, and the cost of noticing in minute one versus minute forty is the whole difference between a useful run and a wasted one.

How do saved logins work, and where do you delete them?

In settings, under the cloud browser section. Manus's documentation describes viewing and managing signed-in accounts at Settings, then Cloud Browser, then Logged-in Accounts, where you can see active sessions and either sign out of a specific account or clear all sessions at once.

Its help centre adds that you can manage saved login information for individual websites in the same place, and that you can turn the whole save-my-login behaviour off from Settings and Cloud Browser if you would rather approve every sign-in by hand.

On the security side, Manus states that browser sessions are encrypted, that each user gets an isolated environment, and that it does not store your passwords. That is the vendor's description of their own design, so treat it the way you would treat any such statement: useful context, not a substitute for your own judgement about which accounts you are willing to sign into on someone else's machine.

When should you use your own browser instead?

When the site is either sensitive or hostile to automation. Manus's own documentation is candid that the cloud browser uses data centre IP addresses, which can trigger more verification challenges, and that some services flag the activity as automated.

The documented alternative runs in your browser instead. Manus describes it as using your residential IP address and your existing sessions, and recommends it for security-sensitive sites such as banking and for services that block data centre traffic outright. Its Browser Operator page describes a Chrome extension installed from the Chrome Web Store and switched on from the Manus connector tab, operating in your local browser context and using your existing signed-in sessions.

That page also states Manus will not see your passwords when working this way. It is worth reading the extension's permission prompt carefully all the same, because an agent operating inside your live browser has a wider reach than one working in a sandbox.

What will not be fixed by taking over?

Content you do not have the right to read. MIT Technology Review's hands-on test found Manus repeatedly hitting CAPTCHA blocks and paywalled articles, with sites citing suspicious activity, and noted that many sites still blocked it even after the reviewer took over to clear the challenge.

The same review noted the version tested did not proactively ask for help when it hit a restriction, and reported crashes and load messages during busy periods. Reviews age quickly with a product that ships this often, so treat those as the shape of the failure mode rather than a current bug list.

Paywalls stay paywalls. If you cannot read the article, the agent cannot either. Give it sources you can actually reach, or take the excerpt yourself and hand it over.
Repeated challenges are a signal. If one site keeps stopping the run, switching to the local browser option is a better move than taking over the same CAPTCHA five times.
Terms of service still apply. A site that flags the traffic as automated is telling you something about what it permits. Taking over does not change that, and it is worth checking before you build a routine on top of it.

None of this makes browser automation a bad idea. It makes it a supervised one, and that is a fair description of where the whole category currently is. The question worth asking is how often you want to be the supervisor.

That question is what shaped Sistava. Instead of driving an agent through a browser and stepping in when it hits a wall, you hire an AI Employee that reaches your tools through proper connections rather than by pretending to be you at a login screen, and that comes back to ask you a question when it is genuinely blocked rather than waiting for you to notice.

What should you set up before a long browsing task?

Sign in first, then start the task. If you know the run needs three sites that require accounts, take over once at the beginning and log into all three rather than being interrupted three separate times, then let Manus save those sessions.

It is also worth checking the takeover notification settings. Manus's cloud browser documentation lists takeover notifications alongside logged-in accounts, session management, browser history and security settings in the same area, which means you can decide how you find out that the agent is waiting on you.

FAQ

How do I take control of the Manus browser?

Manus's help centre says the browser control is in the upper right corner of the running task. Clicking it lets you manually take over the browser, sign in or clear a CAPTCHA, and then hand control back so the task continues.

What does Manus do when it hits a CAPTCHA?

Its cloud browser documentation says complex verifications such as CAPTCHAs, SMS codes and multi-factor prompts trigger a take over request, where you temporarily control the browser to complete the verification and then return control to Manus.

Does Manus remember my logins between tasks?

Only if you let it. Manus's help centre states it asks you to confirm whether to save the login status for a website so it can access that site independently in future. You can manage those saved sessions, or disable the behaviour entirely, in Settings under Cloud Browser.

Where do I see which accounts Manus is signed into?

Manus documents a logged-in accounts view at Settings, then Cloud Browser, then Logged-in Accounts, showing active sessions with the option to sign out of a specific account or clear all sessions at once.

Why does Manus get blocked on some sites but not others?

Manus's own documentation says the cloud browser uses data centre IP addresses, which can trigger more verification challenges, and that some services flag the activity as automated. It recommends its local browser option, which uses your own IP address and sessions, for those cases.

Can Manus get past a paywall?

No, and you should not expect it to. MIT Technology Review's hands-on test reported repeated CAPTCHA blocks and paywalled sources the agent could not reach, with some sites continuing to block it even after a human cleared the challenge. Supply content you have access to instead.

Sources

Take over early, sign in once at the start rather than three times in the middle, and review your logged-in accounts when the job is done. A browsing agent is at its best when you treat the login moments as your part of the work rather than as something it failed to do.