The things people ask us

No. If a person on your team can do the task in a browser, they can teach TaskMate to do it. The people who get the most out of it are usually the ones who know the process, not the ones who know the code.

A simple one — a lookup on a portal, a couple of fields into another system — usually takes under an hour, most of it spent deciding what you want rather than building it. Something spanning four systems with validation rules is more like an afternoon.

Windows and macOS today. Linux is on the roadmap — tell us if you need it and we'll factor that into the priority.

The run stops and tells you which step broke rather than guessing and doing something wrong. You can then re-record or re-describe just that step — the rest of the task stays as it is.

Because it works with the things that live on your computer: your real Chrome browser, with the sessions you're already logged into, and your local files. Running on your machine means your credentials, documents and portal sessions never have to be uploaded to us to get the job done. That's the trade-off for the desktop install — and it's the reason the install exists.

A fair question, and one we'd ask too. Every TaskMate release is code-signed, and we publish the SHA-256 checksum of each installer next to the download so you can verify what you downloaded is what we built. The app is distributed only from taskmate.deccanlogic.com — if you find it anywhere else, it isn't ours. If your IT team wants to review it before rollout, write to us and we'll walk them through it.

Your credentials, browser sessions and files stay on your machine. Tasks execute locally. What reaches our servers is limited to your account, your plan and credit usage, and the run metadata needed to show your history. When you deliberately use an AI-assisted step, only the content needed for that step is sent to the model provider for that step.

No. TaskMate drives the browser session you're already logged into on your own machine. Where a task does need to enter a credential, it's held in your operating system's credential store, not on our servers.

1 credit ≈ $0.01. A deterministic action costs 0.02 credits, so a task with 30 steps costs about 0.6 credits — well under a cent. The 150 free credits each month cover a great deal of ordinary use. AI-assisted steps are billed per 1,000 tokens, and you'll see both kinds broken out separately in your usage view.

The free allowance resets each billing cycle rather than accumulating. Top-up credits on paid plans don't expire at the end of a cycle.

We're setting them with our first users rather than guessing in a spreadsheet. If you're running enough volume to need one, talk to us and we'll work it out together — early users get to keep whatever we agree.

You drop back to the free plan. Your tools and tasks stay where they are; you just get the free allowance of credits each month rather than the paid one.

No — and the difference is the whole point. A model helps you build the tool: it interprets your description, finds the right elements on the page and drafts the steps. Once that tool is saved, the model steps out. Runs are plain deterministic execution, which is why they're fast, cheap and identical each time. You can opt an individual step back into AI when a task genuinely needs judgement.

MCP is an open standard for describing a tool an AI system can call — its inputs, its outputs, what it does. TaskMate builds your recordings and descriptions as MCP tools, which means they compose with the growing ecosystem of third-party MCP tools instead of only working inside our product.

Then use it — either directly, or through an existing MCP tool, and chain that into your task alongside the recorded ones. Recording a browser is what you do when there's no API, not something we think is better than one.

A task can pause and hand control back to you for exactly that step, then continue once you're through. You do the ten seconds that need a human; TaskMate does the rest.