A few
good questions.

Installation & updates

Where do I get the app?

Get native macOS, Windows and Linux installers from the download section. The current published release is v0.9.0-beta.1.

Do I need to build it myself?

No. The installers need nothing else first. Running from source is there for development. Follow Getting started for both.

Why does my computer warn me about it?

The macOS and Windows builds aren't signed. The installation notes show how to open Asyncat the first time.

How do updates work?

Inside the app. Installed builds check GitHub Releases after launch and every six hours, and the profile menu shows Update available when there is one. Asyncat downloads it, checks it against the release checksums, and installs it, or saves it to Downloads for the last step on macOS and with the .deb. Your chats and settings stay. See Updates.

Models & cost

Is Asyncat free?

Yes. The app is free and open source under the MIT license. Cloud models, and voice or image services, bill you separately through your own key. Local models use your own hardware and cost nothing to use. Models → Usage shows requests and tokens per model.

Which model should I use?

One that fits your hardware or provider account and handles tools well. The first-run setup offers a local model that fits this computer, with its download size. The provider guide explains the options.

Can different chats use different models?

Yes. The model menu under the message box changes the chat you're in, and new chats start with the model you picked last. See A model per chat.

Data & connectivity

Can I work offline?

Yes. Every feature works with local models once they are downloaded. Web browsing, connectors, downloads and cloud models need the network.

Does my data stay on my computer?

Yes, except what you send to an online model or service. Installed builds also check GitHub Releases for updates. Files → System shows where everything lives, and Settings → Privacy & safety exports everything to one zip or deletes it all.

What is the command sandbox?

Commands the agent runs go through your operating system's own sandbox, bubblewrap on Linux and sandbox-exec on macOS. They can change files only in the project and can't read your sign-ins and keys. If the sandbox can't start, every command asks first. Windows has no command sandbox yet. See Privacy & safety.

How should I back up my work?

Use Export everything in Settings → Privacy & safety, and keep your own backups of important project folders. Updates never touch your chats, settings or project folders.

Cats & agents

What is a Cat?

An agent that works on its own. A Cat has a mission, wakes on a schedule or when something happens, works in its own folder, and reports in its own chat and in your inbox. You choose how free it is: ask first, ask for risky steps, or work alone. See Cats.

What happened to my scheduled tasks?

They were turned into Cats the first time this version started.

How is an agent different from a Cat?

An agent answers when you pick it in a chat, with its own instructions, files, abilities and model. A Cat works without you. A Cat can use one of your agents as its brain. See Your own agents.

Can I use Claude Code or Codex?

Yes, if they are installed and signed in on this computer. Pick one under the message box, along with its model. See Coding agents.

Making it yours

Can I add tools or skills?

Yes. Add connectors in Settings → Connections → Connectors, and load or write skills in Customize → Tools & Skills. Settings → Personalization tells Asyncat about you and how to answer.

Can I bring my ChatGPT or Claude chats?

Yes. Import a data export in Settings → Privacy & safety. See Bring your chats.

Where can I report a problem?

Use the project issue tracker. Include your app version, operating system, what you expected, and what happened. Remove private information from logs before sharing.