Can Manus run your job search? What it can't do

What did Manus just raise, and does it change your job search?

Manus raised the money, not your interview rate. On 8 October 2026, parent company Butterfly Effect announced a funding round of more than $500 million, led by Boyu Capital and IDG Capital, with existing backers Tencent, HSG — the firm formerly known as Sequoia China — and ZhenFund following on. It's the company's first raise since Meta's acquisition of Manus was unwound at the direction of Chinese regulators, and it leaves Manus one of the most heavily funded general-purpose agent companies anywhere. None of that changes what happens when you hand it your CV. A better-capitalised agent still writes from whatever you feed it, and most job seekers feed it a document nobody has ever scored.

Here's why the round matters to you anyway: money buys reach, and reach means more people will try running a job search through an autonomous agent this quarter than last. Manus isn't a chat window. It plans and executes multi-step work on a cloud computer or inside your own browser, handles files and spreadsheets, builds slide decks, and runs on a schedule while you sleep. The company publishes ready-made playbooks for exactly our use case — a résumé builder that takes your background plus a target role, and a cover-letter generator that researches the employer before it writes a word. Those are real capabilities, not marketing.

So the honest verdict, upfront: Manus is a strong researcher and a decent first-draft writer, and a weak applicant. It's good at the parts of a job search that involve reading the public internet and turning it into structured text. It's unreliable at the two parts that decide whether you get an interview — pressing submit inside a platform designed to stop bots, and stating facts about your own career that will survive a reference check. Treat it like a sharp intern on their first week: fast, eager, occasionally confident about things that aren't true, and in need of a brief you've actually prepared.

One reviewer's comparison stuck with me: Manus behaves like a highly capable intern who sometimes misreads the instructions and makes assumptions to fill the gap. That's a fine description of an agent writing your employment history from a CV that's missing numbers. It won't leave the gap blank. It'll guess, plausibly, in your voice.

Can Manus actually apply to jobs on LinkedIn and Indeed?

No — not in any way you should rely on. Manus can browse, but LinkedIn and Indeed keep their job pages and apply buttons behind a login wall, and both platforms actively defend against automated sessions. LinkedIn's user agreement prohibits bots, scrapers and software that simulates or impersonates human activity, and enforcement has teeth: the company has cut off automation vendors outright, and individual accounts showing repetitive, high-volume behaviour get throttled first and explained to later. The quiet penalties sting more than the loud ones — reduced recruiter visibility, lower application priority, a job feed that stops being relevant — and nobody sends you a notification when it happens.

Easy Apply looks like the obvious target, and that's the trap. It's three clicks, so it seems automatable, but every one of those clicks happens inside your authenticated session. If you point an agent at your own logged-in browser, the platform isn't blocking a robot in the abstract — it's watching your account produce machine-shaped behaviour. Corporate portals are no friendlier. Workday and Greenhouse want an account, an email confirmation, uploaded files, sometimes a CAPTCHA, and a set of knockout questions whose wording changes per employer. I've walked through why auto-apply agents stall at the login wall in more detail, and the conclusion hasn't moved.

The workflow that does hold up is unglamorous and takes about ninety seconds: copy the job ad text out of the browser yourself, paste it into the agent, let it do the thinking, then submit the application with your own hands. You keep the account safe, the agent keeps doing what it's genuinely good at, and you read the thing before it represents you. Also worth saying plainly: paste the text rather than the link. LinkedIn and Indeed block fetching, so handing any tool a job URL from those sites usually returns an error page or, worse, a confident summary of nothing.

What does Manus genuinely do well in a job search?

Research and drafting — and it's properly good at both. Give Manus a pasted job ad and it will separate the hard requirements from the nice-to-haves, name the tools and certifications the employer keeps repeating, and flag the phrasing that's clearly been lifted from an internal competency framework. Point it at an employer and it reads the website, recent press, product pages and leadership profiles, then hands you a briefing that's genuinely useful before a first-round call. Ask it for twenty likely interview questions for a specific role at a specific company and you'll get a list worth rehearsing. That's hours of browsing compressed into one unsupervised run.

The scheduling is the underrated part. Because Manus runs asynchronously in the cloud rather than waiting on you to prompt each step, you can set a recurring task — scan a saved list of company career pages each morning, pull anything matching your criteria, draft an opening paragraph for each — and review the output over coffee. Costs are metered in credits that burn by the minute, and a deep research run can swallow a free plan's daily allowance in one go, so this is a paid-tier habit. Budget it like a tool, not a toy: one well-aimed run beats twenty speculative ones.

Where it earns its keep is breadth, not judgement. An agent that reads forty company blogs and extracts a pattern is doing something you'd never do by hand. An agent writing the line "reduced deployment time by 40%" about your last job is doing something only you can verify. Split the work along that line and the whole thing becomes useful. The same split applies to the chat-style assistants people now run their searches through — the division of labour between a messaging-based agent and a file-based CV tool follows exactly the same logic.

Where do AI agents break your CV?

Three places, every time: invented achievements, one generic CV recycled across every application, and a layout that looks sharp and parses badly. The first is the dangerous one. When your CV says "managed the migration" with no scale attached, an agent asked to make it sound impressive will attach a scale. You'll get "led migration of 120 microservices, cutting infrastructure costs 30%" — specific, confident, and entirely fabricated. It reads beautifully until a hiring manager asks which 120 services, or a referee is asked to confirm the saving. Nobody set out to lie; the agent was simply never told the real number, so it supplied one.

The second failure is subtler and more common. Agents are good at producing output, which makes it tempting to generate one polished master CV and send it everywhere. That's the opposite of what tailoring means. A CV aimed at a job ad mirrors that ad's vocabulary — if they wrote "stakeholder management" and you wrote "working with the business", the keyword match isn't there, regardless of how similar the work was. Then there's the tell: a certain register of agent prose — "spearheaded", "leveraged", "results-driven professional" — that recruiters now spot in seconds. If you're unsure how yours reads, it's worth checking whether your CV sounds AI-written before you send it.

The third is purely mechanical. Agents export handsome PDFs, and handsome often means two columns, icons for contact details, skills in text boxes and a sidebar for certifications. Parsers read in a single stream. A two-column layout can interleave your job titles with your skills list, icons carry no text at all, and content inside a text box is sometimes skipped entirely. You end up with a document that reads perfectly to a human and arrives at the recruiter with your employment dates scrambled. No agent will tell you this, because the agent can't see what the parser saw.

How do you give an agent a base CV worth tailoring?

Score and fix your CV before any agent touches it — that's the whole trick. Run it through a free CV analysis first and you'll get an overall score out of 100, five category scores covering experience, tech stack, impact and ownership, clarity and structure, and ATS compatibility, plus a visual layout analysis that shows how your formatting actually holds up. Upload a PDF, Word file or even a photo of an old printout — Hebrew documents are read by OCR too. Now you know which numbers are missing and which sections are thin, which means you can supply the real figures instead of letting an agent guess.

With a scored base CV in hand, the targeting step is where agents and a proper CV tool stop competing and start stacking. Paste the job description text into the analysis and you get the skills your CV is missing against that specific ad, suggestions tailored to it, and ATS-friendly rewritten versions across six templates that mirror the job's own vocabulary. If you'd rather rebuild from scratch, the chat-based CV builder takes your old document or a plain description of what you've done and refines it by conversation, exporting to PDF or Word — and Hebrew CVs are written in Hebrew and exported right-to-left.

Here's my one contrarian take: volume is the wrong thing to automate. The appeal of an agent is firing off eighty applications while you sleep, and eighty untailored applications produce roughly the same result as none — except you've also risked your LinkedIn account. Automate the research, the drafting and the admin. Keep the choosing and the sending. Twelve properly tailored applications a week, each built from a CV you've scored and each submitted by hand, beats a scheduled firehose. The agents that help are the ones you brief well, and a brief is just a good CV plus a real job ad.

Frequently asked questions