How to Use Manus: The Beginner's Guide (Chat Mode, Agent Mode, and Credits)

CodexFuse Team · 2026-10-29T05:02:09.362Z · 19 min read

Manus finishes whole tasks instead of just answering questions. This beginner guide covers Chat vs Agent mode, your first task brief, credits math, pricing, and Academy.

Learning how to use Manus starts with one shift in thinking: you do not chat with Manus the way you chat with a chatbot, you hand it a whole task and let it finish the job. Manus is an autonomous AI agent that plans a multi-step job, works through it in its own cloud computer, and hands back a finished result — a report, a slide deck, a spreadsheet, or a working website. This Manus AI tutorial walks through the full beginner path: signing up, choosing between Chat Mode and Agent Mode, writing your first task brief, understanding Manus credits, and checking the result before anyone else sees it.

If you already use fixed automation, the difference is worth naming plainly. A tool like Zapier runs steps you defined in advance, while Manus decides the steps itself from your goal. The n8n beginner guide covers the self-hosted middle ground, where you build the workflow and an agent node works inside it. Manus sits at the other end: you describe the outcome, and it figures out the middle.

How Do You Use Manus?

To use Manus, sign up at manus.im, pick Chat Mode for a quick question or Agent Mode for a finished deliverable, write a brief that names the goal, the scope, and the exact output, let Manus plan and work, then check every fact and file before you use or share the result.

1. Sign up at manus.im and open a new task from the main task box.
2. Pick your mode first. Use Chat Mode for a fast answer and Agent Mode when you want a finished file, report, deck, or site.
3. Write a brief, not a question. Name the topic, the scope, the deliverable, the format, and any limits on length or sources.
4. Watch it work, and steer it. Follow the plan as it runs, answer when it asks, and pause or redirect it if it drifts off brief.
5. Verify before you trust it. Open the files, spot-check the numbers against their sources, and fix anything wrong before publishing.
6. Save the brief that worked. Turn a repeating job into a project so you never write the same instructions twice.

That six-step loop is the whole system. Everything else in this guide — modes, projects, credits, Academy, and safety — exists to make one of those steps cheaper, safer, or repeatable.

What Is Manus, and How Is It Different From a Chatbot?

Manus is a general-purpose AI agent, not a chat window with a longer memory. A chatbot answers your message and waits; Manus takes your goal, breaks it into steps, browses the web, runs code, edits files, and keeps going until it has something to hand back. Its own documentation describes it as an autonomous agent with its own virtual computer, internet access, and file system, and that description matches how it behaves in practice.

The practical difference shows up in what you ask for. Ask a chatbot a question and you get an answer. Ask Manus for a deliverable and you get a thing: a competitor report as a PDF, a slide deck with speaker notes, a spreadsheet built from scattered web pages, or a small website you can publish. That is why the brief matters so much more here than in chat. A vague goal produces a long, expensive wander; a tight brief produces a short, checkable job.

Manus for beginners is therefore less about learning buttons and more about learning to delegate. You are briefing a fast, literal, tireless colleague who will do exactly what you wrote, including the parts you left ambiguous. Write the brief you would give a freelancer on their first day, and Manus will behave much better than it does with a one-line question.

Chat Mode vs Agent Mode: Which One Should You Pick?

Pick Chat Mode for quick questions and Agent Mode for finished work. Chat Mode is the cheap, fast lane: ask something, get an answer, move on. Agent Mode is the full machine: Manus plans the job, opens its virtual computer, browses, runs code, and builds the deliverable. The two modes answer different needs, and mixing them up is the most common beginner mistake.

Use Chat Mode when the answer itself is the product. Examples include explaining a concept, drafting a short email, summarizing a pasted paragraph, or checking a definition. Use Agent Mode when the product is a file or a multi-step result: research across many sources, a deck, a data table built from the web, a website, or a report with charts. If the job needs more than one tool — browsing plus files plus code — it belongs in Agent Mode.

The money rule follows directly from that table. A question that Chat Mode answers in seconds should never be sent through Agent Mode, because Agent Mode pays the full cost of planning, browsing, and executing for a job a single reply could have settled. Start in Chat Mode by default, and promote a job to Agent Mode only when you can name the deliverable.

How Do You Write Your First Manus Task?

Your first Manus task should be small, real, and easy to check. Pick a job you already understand, so you can tell a good result from a confident-looking bad one. A five-source research note, a one-page summary of a long document, or a simple comparison table all work. A 40-page market report with money riding on it does not — not for a first run.

A strong brief has five parts, and the order matters. First, the goal in one sentence. Second, the scope: what is in, what is out, and which sources or regions count. Third, the deliverable and its format, such as a two-page PDF with a summary table at the top. Fourth, the limits: length, number of items, and anything it must not do. Fifth, the check: what a good result looks like, so you can judge it in one minute.

Here is a before-and-after that shows the difference. A weak brief says: “Research AI.” A strong brief says: “Research the five most common questions small accounting firms in the UK ask about switching to cloud bookkeeping. For each, give a short plain-English answer and one example. Deliver it as a two-page PDF with a summary table at the top.” The second brief is checkable, and checkable briefs cost fewer credits because Manus wastes fewer steps guessing what you meant.

While the task runs, watch the plan. Manus shows its steps as it works, and you can step in when it heads somewhere useless — a wrong source, an oversized scope, a format you did not ask for. Pausing early to correct course is almost always cheaper than letting a bad run finish and starting over. When it finishes, download the files and read them before you forward anything.

What Can Manus Actually Build for You?

Manus can build research reports, slide decks, spreadsheets, data summaries, and websites, plus the in-between jobs like meeting notes and email-started tasks. Its documented capabilities include Wide Research, which splits a large research job across several agents running in parallel; Manus Slides, which produces a complete presentation with research, visuals, and speaker notes; data analysis that turns raw material into charts and insights; and a website builder that edits with plain-language changes and can publish the result.

Two features change how beginners should organize their work. Projects hold a master instruction and a knowledge base of files, and every new task inside the project inherits both — so a weekly report stops needing the same setup paragraph every time. Mail Manus lets you start a task by forwarding an email to your personal bot address, with only approved sender addresses able to trigger tasks; it is handy for turning a long supplier email into a summary or a draft quote outline.

Manus also works where your own logins live, and that needs care rather than enthusiasm. Its Browser Operator feature uses your local browser, with each session authorized by you, and closes when you close the tab. Its desktop app can read and write folders you allow, with each command needing approval. Keep both on a short leash at first: use them only for sites and folders the task genuinely needs, and never hand over a main inbox or drive before you have tested with a throwaway account.

Manus Credits Explained: How Much Does a Task Cost?

Manus credits are the currency every task spends, and a task's cost depends on how much it browses, reads, writes, and computes. Manus does not publish a fixed price per task, and no pre-task estimate you can rely on exists in its public pages, so the honest way to budget is to measure your own tasks. Note the credits before and after each run for your first week, and you will quickly learn what your typical jobs cost.

The plan figures below come from third-party checks of Manus's own pages in mid-October 2026, because the vendor pricing page itself rendered without extractable numbers in this guide's check today. Treat every figure as “check again at sign-up”: Manus changes its pricing pages often, and one September 2026 check found the public page showing paid tiers only, with no free plan card displayed. The free plan described below is widely reported in October 2026 guides, but it is not guaranteed to be offered to every new account on the day you sign up.

Three rules shape everything else about credits. Daily refresh credits reset to 300 at 00:00 UTC and unused daily credits are removed, so logging in daily matters on the free plan. Monthly credits do not roll over to the next cycle, except that separately purchased add-on packs are treated differently and Manus's own pages disagree on their expiry — plan around the stricter reading. Tasks that fail because of a technical issue on Manus's side get their credits refunded, and storing finished work costs nothing.

The original arithmetic below turns those plan figures into a budgeting rule. It uses assumed task sizes, not measured Manus prices, because Manus publishes none — the point is the method, and you should replace the assumptions with your own measured costs after week one.

Read that table as a warning against the most expensive beginner habit: sending big, vague jobs to Agent Mode. At $5 per 1,000 credits on every paid tier, the tiers buy capacity, not a discount — a bigger plan does not make your tasks cheaper, it only lets you run more of them. Start on the free tier if it is offered, measure a week of real tasks, and only then pick the tier your measured usage justifies.

What Is Manus Academy, and Should Beginners Take It?

Manus Academy is Manus's free learning platform for applying AI agents to real work, and beginners should take its Fundamentals course before spending serious credits. The Academy is separate from the Manus platform itself: the courses are free, while running agent tasks on the platform uses the credit plans above. Its Fundamentals course covers platform navigation, interface basics, prompt fundamentals, and how to structure goals for autonomous execution, in roughly two to three hours.

The Academy's value is practical rather than theoretical. Its courses progress from fundamentals to role-specific intermediate and advanced modules — business analysis, product management, finance, and team administration among them — with documented use cases and templates drawn from real users. Completing courses earns credentials for practical capability in your domain, and early learners also received platform credits as a practice incentive, though that offer should not be assumed still open.

Use the Academy as your practice ground, not your destination. Take Fundamentals, then run one small real task from your own work the same day. The course teaches you how Manus thinks; only a real task with a checkable result teaches you what your briefs cost and where your checking habit needs to be strict.

Manus vs ChatGPT: Which One Do You Actually Need?

Choose ChatGPT when you want a conversation and Manus when you want a finished job handed back. ChatGPT answers, explains, drafts, and iterates with you in the same thread, and its agent features work inside that conversation. Manus plans and executes on its own, in its own computer, while you do something else. The ChatGPT beginner guide covers the chat-first path, and the Google Gemini beginner guide covers the other major chat-first alternative.

The honest comparison is about the shape of the work. If the job is thinking with you — brainstorming, refining an argument, drafting and redrafting — a chatbot is the better tool, because its back-and-forth is the point. If the job is doing without you — gathering from fifty pages, building the deck, assembling the table — Manus is the better tool, because its independence is the point. Many people end up using both: chat to think the brief through, then Manus to execute the finished brief.

Cost makes the same distinction. A chat subscription is a flat monthly bill for conversation, while Manus is a credit meter that rises with every browse, read, and write. If your work is mostly questions, Manus's meter is the wrong shape of bill. If your work is mostly deliverables, the meter is fair — you pay for finished work, not for time spent chatting about it.

Who Owns Manus, and What Happened to Your Data?

Manus is owned by Butterfly Effect, and it is independent today after a Meta acquisition deal was announced and then blocked. Meta announced the purchase in December 2025, in a deal reported at more than $2 billion. On April 27, 2026, China's National Development and Reform Commission asked the parties to withdraw the deal, saying it had decided to prohibit foreign investment in the Manus project. On August 11, 2026, Manus told users it would return to operating as an independent company.

That independence came with a data consequence you should know about. Manus said data some users generated on or after December 29, 2025 would be deleted on August 23 and 24, 2026, unless backed up. If you used Manus between the deal announcement and that August window, check whether your account fell into the deletion group before assuming old tasks are still there. On October 8, 2026, Butterfly Effect announced a new funding round of more than $500 million, led by Boyu Capital and IDG Capital, with Tencent, HSG, and ZhenFund following on.

None of that changes how you use Manus day to day, but it does change what to read before connecting anything sensitive. Read the terms and privacy policy you accept at sign-up, and treat ownership history as one more reason for the safety rules in the next section — not as a reason to panic, and not as something to ignore.

How Do You Use Manus Safely?

Use Manus safely by starting with no connectors, adding one at a time only when a task needs it, and never pasting passwords, client secrets, or private data into a task. Manus's sandbox protects your laptop — the agent's code runs in its cloud machine, not on your computer — but once you connect Gmail, Drive, or a CRM, that machine holds live sessions to those accounts. Least privilege, applied by hand, is the whole safety system.

Four habits cover most of the risk. First, give the agent its own accounts where possible: a throwaway mailbox and a test folder beat logging your main accounts into its cloud browser. Second, prefer approval for anything that writes — the desktop app asks you to allow each local command, and “once” is the right answer until a command has earned “always.” Third, save Browser Operator for sites that genuinely need your real login, authorize each session yourself, and close the tab to stop it instantly. Fourth, never join a shared project from a stranger: a September 2026 security report showed a project owner's instruction field loading as trusted configuration into members' agents, and that single click became code execution in another member's sandbox.

Finally, remember that Manus's own help center has a section on hallucinations, and its authors are right to include one. An agent that browses, reads, and writes at machine speed can also misread, miscopy, and confidently invent at machine speed. Check every number, quote, and claim before it reaches a client, a boss, or the public. Verification is not an optional final step in agent work — it is the step that makes the rest of the work usable.

What Should Your First Week With Manus Look Like?

Your first week with Manus should be one small real task a day, measured, not a heroic first project. Day one: sign up, take the Academy Fundamentals course, and run one tiny Chat Mode question. Day two: run your first Agent Mode task — the small, checkable job from this guide — and write down the credits before and after. Day three: turn it into a project if it repeats, with the master instruction and the files it always needs.

By day five you should know three numbers: what your typical task costs, how many tasks your free daily credits actually cover, and how long checking takes you. Those three numbers decide your plan, not any pricing page. If checking takes longer than doing the job yourself did, your briefs are too vague — tighten the scope and the deliverable before you spend another credit. If the free daily credits cover your real workload, stay free until they provably do not.

End the week by deciding what Manus is for in your work. The Zapier beginner guide covers the fixed-workflow branch for jobs whose steps never change, and this guide covers the open-ended branch for jobs whose steps you cannot fully predict. Most people keep both: fixed workflows for the routine, and an agent for the research, the decks, and the one-off builds that used to eat whole afternoons.

Sources

Manus pricing and credit rules in this guide follow third-party checks of Manus's own pages in mid-October 2026, with the vendor page's own rendering noted where it could not be read: Lilach Bullock, “How to Use Manus AI: Tasks, Projects and Credits” (updated October 24, 2026, https://www.lilachbullock.com/how-to-use-manus-ai/) — free-plan rules (300 daily credits, reset at 00:00 UTC, 1,500 monthly cap, Manus 2.0 Lite only), credit spending order, and the $20 plan's 4,000 monthly credits; Jahanzaib, “Manus AI Agent: Owner, Cost, Risks and Alternatives” (prices read from vendor pages October 17, 2026, https://www.jahanzaib.ai/blog/manus-ai-agent-what-it-does-pricing-alternatives) — ownership dates (December 2025 Meta deal, April 27, 2026 block, August 11, 2026 independence, August 23–24, 2026 data deletion, October 8, 2026 funding round), the plan table, the $5-per-1,000-credits arithmetic, and the add-on credit expiry disagreement; Boost with AI, “What Is Manus AI and Its Academy? A Beginner's Guide” (updated October 27, 2026, https://www.boostwithai.se/manus-academy-beginners-guide/) — Academy structure, Fundamentals course length, and plan figures; Manus, “Introducing Manus Browser Operator” (https://manus.im/blog/manus-browser-operator) — per-session authorization, logged actions, and closing the tab to stop; Manus pricing page (https://manus.im/pricing) — checked October 29, 2026, rendered without extractable plan numbers, which is why plan figures here are attributed to the dated third-party checks above rather than presented as a live vendor read; Ailect, “Manus AI Review” (September 2026 pricing check, https://ailect.com/manus-ai-review-in-depth-testing-pricing-verdict-2026/) — the finding that the public pricing page then showed paid tiers only, and observed task-size credit ranges.

Frequently Asked Questions

Is Manus AI free?

Manus is widely reported to offer a free plan with 300 credits a day when you log in, capped at 1,500 a month for free users, usable only with Manus 2.0 Lite — but check at sign-up, because one September 2026 check found the public pricing page showing paid tiers only. If no free plan is offered to your new account, budget $20 a month for the entry Pro plan.

What is the difference between Manus Chat Mode and Agent Mode?

Chat Mode gives you a fast answer in the conversation and costs far fewer credits. Agent Mode plans a multi-step job, works in its own virtual computer, and hands back finished files such as reports, decks, spreadsheets, or websites. Use Chat Mode for questions and Agent Mode for deliverables.

How many credits does a Manus task use?

It depends on how much the task browses, reads, writes, and computes, and Manus publishes no fixed price per task. As a budgeting guide only, a short lookup may use around 100 credits, a focused research note around 300, and a deep research run or small website around 1,000 — measure your own tasks, because your briefs will cost differently from anyone else's.

Do Manus credits roll over?

No. Daily refresh credits reset to 300 at 00:00 UTC and unused daily credits are removed, and monthly plan credits do not roll over to the next cycle. Separately purchased add-on credits are treated differently, and Manus's own pages disagree on their expiry, so plan around the stricter reading while you hold a paid plan.

Is Manus safe to connect to my email and files?

Connect as little as possible at first. Start with no connectors, add one at a time only when a task needs it, prefer approval for anything that writes, and test with a throwaway account before connecting a main inbox or drive. Never paste passwords, client secrets, or private data into a task, and always verify outputs before sharing them.