QUICKWINS.AI/TOOLS/AI-ML/MANUS

Manus

BUILT-IN AI INVOKE-ONLY
← AI and ML
6 of 14Reads only
YESBuilt-in AI
14Actions
0Event triggers

AI-powered task automation and workflow management platform

What your AI can do here Connector snapshot 2026-08-23

Connect Manus to ChatGPT or Claude with your own Manus login or an API key, and your AI can run 14 of its actions. 6 of those only read and change nothing. 3 can delete or send something that cannot be undone, so those need a boundary you set before anyone runs them.

This describes what Manus supports, which is one half of the picture. Where you start depends on the task and on how much you have handed over before, and no page can tell you that from a count.

Things you could hand over

Checked 2026-08-22

Candidates, not recommendations. These are things Manus supports once connected. Whether any of them is right for you depends on your own work, which this page knows nothing about.

What did I set running in Manus, and where did it get to? No prior practice needed

Your assistant can list your Manus tasks, filter them by status, project, search term or time range, and then read any single task in full: its current status, the output messages it has produced, its metadata, and the credits it has consumed. That turns a check-in from opening the Manus interface and scrolling into asking a question and getting the state of everything in one answer. It reads only.

What you end up with. A current status list of your Manus tasks, with the finished ones summarized and the stalled ones named.

Start with it bringing you the information and you deciding.

Never without. Read a task's own output as its claim about what it did, not as verified fact. Status tells you the run finished, and it does not tell you the result is correct.

Built on: List tasks / Get Task

What is Manus actually costing us this month? Assumes some practice with AI

Credit usage is on the task record, so the assistant can pull the tasks for a period and total the credits by project or by the kind of work. The number that matters is usually not the total. It is the one task that ran four times because the instructions were vague, and that only shows up when you look at the tasks side by side.

What you end up with. Credit consumption broken out by project and task, with the expensive repeats identified by name.

Start with it bringing you the information and you deciding.

Never without. Treat this as the account's own usage record, not as a bill. Reconcile it against your actual invoice before you use it to make a spending decision.

Built on: List tasks / Get Task

Handing a long job off to Manus without leaving what I am doing Assumes some practice with AI

It can create a Manus task from a description you give it, attach files, and specify which connectors the task is allowed to use, then hand back the task ID. Because the same action continues an existing task when you pass that ID, a follow-up correction is another instruction rather than a fresh start. The pattern that works is to dispatch, get on with something else, and read the result back later.

What you end up with. A Manus task running against your instructions, with its ID recorded so you can check on it or add to it later.

Start with it doing the work and checking with you.

Never without. Be deliberate about which connectors the task is allowed to use. You are handing a second agent access to the systems you name, and it will act on them while you are not watching.

Built on: Create New Task / Get Task

The same kind of job keeps coming up. Can it stop starting from scratch? Assumes some practice with AI

Manus projects hold default instructions that apply to every task created inside them. So the assistant can create a project once with the standing context, then create each new task against that project ID rather than restating the rules every time. It can also list your existing projects to find the right one. This is the difference between explaining your standards weekly and writing them down once.

What you end up with. A Manus project carrying your standing instructions, with new tasks inheriting them instead of being briefed from scratch.

Start with it doing the work and checking with you.

Never without. Read the default instructions back after they are written. Every future task in that project inherits them, so a vague line becomes a standing vague line rather than a one-time mistake.

Built on: Create New Project / List Projects / Create New Task

Getting the source documents to Manus so the task has something to work from Assumes some practice with AI

Uploading is two steps and the connector only covers one of them. It creates a file record and hands back a presigned upload URL that expires at a stated time. Putting the content at that URL is a separate PUT request that Manus does not expose as an action, so whether it happens depends on what else your assistant can reach. Once content is there, the connector can read the file's status, which moves from pending to uploaded, and list the ten most recent uploads.

What you end up with. A file record created with its upload URL, and the file's status read back so you know whether the content actually arrived.

Start with it doing the work and checking with you.

Never without. Check the status reads uploaded before you attach the file to a task. A record can exist while the content never made it, and the task then runs against nothing without saying so.

Built on: Create File / Get file details / List uploaded files / Create New Task

Sharing a Manus result with someone who is not on the account Assumes you have done this kind of handover before

The assistant can update a task's title, its visibility in the task list, and its sharing settings, including turning public sharing on. That is the route to sending a result to a client or a colleague without adding them to the account. It is also the setting with the widest consequence in this connector, because it changes who can reach the task output.

What you end up with. The task titled clearly and shared exactly as far as you intended, confirmed by reading the setting back.

Start with it doing the work and checking with you.

Never without. Never enable public sharing without reading the full task output first. The output carries whatever the task saw, including anything it pulled from a connected system, and public means reachable by anyone with the link.

Built on: Update Task / Get Task

Clearing out finished tasks and the files they used Assumes you have done this kind of handover before

It can delete a task and delete an uploaded file, and Manus is direct about what that means. Deleting a task is permanent and cannot be undone. Deleting a file removes both the record and the stored content. Run it as two moves. Have it list what matches your rule, read that list yourself, then approve deletion of exactly those items.

What you end up with. The finished tasks and their source files removed, and a saved list of every item deleted with its ID and title.

Start with it doing the work and checking with you.

Never without. Never approve a deletion described only as a rule, and save any task output you still need before deleting. Manus states that deleting a task cannot be undone, and deleting a file removes the stored content as well as the record, so both take the evidence of what was done with them.

Built on: List tasks / List uploaded files / Get file details / Delete task / Delete a file

What we could not establish
  • The connector exposes creating a file record and returning an upload URL, but the actual content upload is a separate PUT to that URL. Whether an assistant can complete that step depends on how it is connected, and that was not confirmed.
  • Nothing in the connector reports account credit balance or plan limits, only per-task credit usage, so it cannot tell you how much budget is left.
  • Manus can post task events to a webhook, and the connector can register one. Nothing here consumes that feed, so it was not written up as a candidate: it needs an HTTPS endpoint you own and a receiver that verifies the RSA signature.
  • No time saving is claimed for any candidate on this page, because none has been measured.

BUILT-IN AI Ships AI features you can turn on yourself, with no developer and no new purchase.

No event triggers. Every win in this app is invoked: you or an agent asks, an action runs.

Capabilities
WRITE Create and update records from an instruction.
READ Pull records, messages, and content out on demand.
WEBHOOK-TRIGGERS Fire automations when events happen in the app.
FILE-TRANSFER Move files in and out of the app.