How to use TabTabTab

TabTabTab is most useful when you treat it the way it's designed instead of as a remote laptop you SSH into. A few rules of thumb make the difference between "this is a slower local setup" and "I'd never go back."

The environment is your laptop. The workspace is the unit of work.

An environment is a long-lived cloud VM. Create one per project (or one per role, like "client work" vs "side project") and keep it around. It's not a task. It holds your shell config, your installed tools, your secrets, and your repos. Destroying it is like wiping your laptop.

A workspace is a git worktree on that VM. It's the unit of work. Make a new one per feature, per bug, per agent experiment. They're cheap. Don't try to share one workspace between two pieces of work.

When you find yourself wanting "a fresh start," create a new workspace, not a new environment.

Keep work isolated, not tangled.

Because each workspace is its own worktree with its own devcontainer, you can split feature work, bug fixes, and experiments without them stepping on each other. Use that. If a task feels risky, create another workspace instead of dirtying your main one.

You'll naturally start treating coding like steering persistent work, not babysitting a laptop.

Use the TabTabTab IDE for agent work.

• TabTabTab IDE for coding work. It is based on our OpenCode fork, sees the repo, edits files, runs commands, and stays attached to the same cloud machine.
• Codex login or provider API keys for models. Codex subscription login works well because it is cheap; OpenRouter, Anthropic, OpenAI, and other supported provider keys work too.
• The browser agent to verify what the coding agent did. It starts your app on its own and confirms the change actually runs.
• VS Code / Cursor / SSH when you want to touch the code yourself. Same VM, same files; you're just attaching a different head.

The intended loop is: agent session makes a change → browser agent verifies it → you review the diff → merge.

Steer, don't babysit.

TabTabTab is built for steering agents, not staring at them. Watch the agent if you want. Jump in to redirect, correct, or push back. But you don't have to. The whole point of running on the VM is that you can close the tab, close your laptop, and pick it up from your phone an hour later, and the agent has kept going either way. Stay close when the work matters; walk away when it doesn't.

Set up secrets once.

The first time you onboard a repo, paste the .env values into Secret Manager. After that, new environments get those values written in for you. Don't re-paste OPENAI_API_KEY every time you spin up a new VM; let Secret Manager handle it.

Move between devices freely.

The dashboard's Open button works on every device. Start work on your laptop, check in from your phone on the bus, finish from a tablet on the couch. There's no syncing step because the workspace lives in one place.

When in doubt, email us.

If a workflow feels awkward, it usually means the product should do more, not that you should work around it. Email support@tabtabtab.ai with what you were trying to do; that's the most direct path to a fix.