← LoomWatch home
GuideFAQ

LoomWatch FAQ. Plain answers.

What people ask before they try LoomWatch. Something missing? Ask on GitHub Discussions.

Is LoomWatch free?

Yes. LoomWatch is free and open source, with no account and no subscription of its own. What costs money is the AI models, and you choose how to pay for them: with the ChatGPT or Claude plan you already have, or with an OpenRouter key that's charged at each model's price.

Do I need an account or an API key?

Not for LoomWatch: it has no account of its own. To get answers you need one way to pay for AI models: your ChatGPT Plus or Pro plan, your Claude Max or Team plan, or an OpenRouter key, which reaches Claude, GPT, Gemini and many more. LoomWatch asks which when you send your first request.

Do I need a ChatGPT, Claude or Gemini subscription?

No, but you can use ChatGPT Plus or Pro, or Claude Max or Team, if you have it. With ChatGPT, click Continue with ChatGPT in LoomWatch and allow it on ChatGPT's own page. Agents on ChatGPT models then use your plan's allowance, at no extra charge, until they reach its limit. Claude Max and Team plans come with API credit every month. You link the plan to the Claude Console, make a key there and paste it into LoomWatch, and agents on Claude models spend that credit. Claude Pro has no API credit, and a Claude sign-in can't be used, because Anthropic doesn't allow it in other apps. Gemini plans can't be used. Without any of these, an OpenRouter key reaches every model.

Do I need to know how to code?

No. You build agents and teams by clicking, and you tell them what to do in plain words. Getting started takes a few minutes.

What does a run cost?

It depends on how each agent's model is paid and how much work the job takes. On your ChatGPT plan, nothing extra: it counts against your plan's limits. On your Claude plan's credit, each call costs its list price, taken from your monthly credit. Through OpenRouter, each call is charged at that model's price. After every run, its receipt shows what it cost, and the Usage page shows what LoomWatch has used on each. If a run costs more than $5.00 and twice what it usually does, it stops and asks you first. You can also give an OpenRouter key a credit limit, so it never spends more than you chose.

Is my work sent anywhere?

LoomWatch runs on your computer and keeps your teams and their history there. What an agent works on (its instructions, your request, and what it reads: files, pages, and your notes, events or Health readings when you gave it those apps) goes to the company whose model it uses, as with any AI service, and through OpenRouter when that's how it's paid. Skills are downloaded only when you add one. A tool you connect, such as Todoist, gets what an agent sends it, and the agent asks you before each action. Nothing else leaves unless you turn on Send to Notion.

Which computers does it run on?

Macs with macOS 11 or newer, Apple silicon or Intel, and PCs with 64-bit Windows 10 or 11.

Can agents in one team use different models?

Yes. Each agent has its own model, and says how it's paid. A team can research with Claude on your Claude credit, check with GPT on your ChatGPT plan and draft with a free model through OpenRouter. Building a team shows how.

Can an agent do things on my computer without asking?

Only what you switched on for it: searching the web, editing files in its own folder, or running commands. Anything else, it stops and asks you in the chat, and if nobody answers within 10 minutes the answer is no. Choose Auto beside the chat's send button and the team gets on without you, changing and deleting files only in the folder each agent works in, and asking before publishing or changing anything outside it. A tool you connected, such as Todoist, asks before each action it takes in your account there, unless you chose Always allow this for that action. In Auto, agents look in the Apple apps you gave them, such as Calendar or Notes, without asking. Reviewing every step explains the choices.

Is it safe to run a team someone shared with me?

Treat a team from someone else with care, as you would any file from the internet. The first time you run it, LoomWatch shows what it will do before anything starts: which model each agent uses, what each may do without asking, which folders it can read, and whether it runs on a schedule. Choose Trust and run only if you trust where it came from.

Can a team run by itself every morning?

Yes. A team can run on a schedule and send its answer to Notion. It runs while LoomWatch is running and the computer is awake, and a review stop still waits for you.

Is LoomWatch a chatbot?

No. It runs teams of agents, each with its own job, and shows you every step. It isn't a cloud service either: it runs on your computer. It doesn't sell AI models: your agents use them through your own ChatGPT plan, Claude credit or OpenRouter key. And it isn't autopilot: you choose where the team stops for you, and it waits.

How do I remove LoomWatch?

On a Mac, open LoomWatch’s menu (point at the black notch at the top middle of the screen and click ⋯; if you turned off Show at the notch, click LoomWatch’s face in the menu bar instead), turn off Start at login, choose Quit LoomWatch, and drag LoomWatch from Applications to the Trash. On Windows, remove it in Settings › Apps › Installed apps. Your teams and their history stay on your computer until you delete them too; the README says where they are.