Launching a Container
Launching a container is a short, deliberate flow: open the launch screen from a project, pick a size that matches your work, choose the desktop editor, pick the AI account the container's assistant uses, and spin it up. This guide walks every choice on that screen and shows what each status badge means as the container comes up, so you know exactly what you are getting and what to expect while it provisions. After this page you will be able to launch a container of the right size, in the right place, and connect to it with confidence.
Open the Launch Screen
After this section you will know where the launch flow lives. Containers launch from a project, because a container is a workspace for the work in that project. Open the project and choose AI Container in its sidebar to reach the Launch a container screen. A short explainer at the top tells you plainly what you are about to get: a cloud environment with a terminal, VS Code, and your files saved to your own private storage, billed only while it runs. Read it once; it is the honest summary of the product.
If AI Container is not in the project's sidebar, containers may be turned off for your organization or for this project. An administrator controls which sizes and backends your team may use; see Container Policies and Administration.
Read a Size Card
After this section you will be able to pick the right size for a task. Sizes are presented as cards, one per tier. Every card carries the same four facts, and choosing well is a matter of reading them:
| On the card | What it tells you |
|---|---|
| Tier name | The named size, such as Standard. Larger tiers have more compute and draw more credits. |
| vCPU and memory | The processing power and RAM the container gets. Match this to your workload: a shell and an editor need little; a build or a browser-driving agent wants more. |
| Credits per hour | The exact rate this size draws from your usage credits while it runs. This is the number that governs cost. |
| Backend and region | Which cloud the container runs in and where, geographically. This matters for latency and for data residency. |
A size is identified by the full combination of its backend, its tier, and its region, so the same tier offered in two regions is two distinct choices. Pick the smallest tier that comfortably runs your work; you are never locked in, and you can launch a larger one for the next task. The rates and exact tiers shown on the cards are the authoritative figures; this guide does not restate specific numbers because your available sizes depend on your plan and your organization's policy.
Which size should I start with? The smallest one that runs your work. A terminal session and light editing need very little. Move up a tier only when a build, a test suite, or an agent workload actually needs more. Since you pay per hour of running time, an oversized container just costs more for the same result.
Choose the Desktop Editor
After this section you will know what the desktop of a container shows when it starts. Every container launches with all three surfaces: a terminal, VS Code, and a graphical Linux desktop you can see and control in the browser. There is nothing to turn on. The Desktop editor choice decides what the desktop opens:
- Standard desktop, the default: a plain Linux desktop with a browser, for desktop applications, a full browser including its chrome, or anything that is not a page or a terminal.
- Windsurf (Devin Desktop): the desktop opens the official Linux Windsurf editor. The terminal's AI assistant is still the account you choose below.
The surfaces themselves are covered in Terminal, VS Code, and Desktop.
Run the In-Container AI on Your Own Account
After this section you will know how the AI assistant inside a container is powered. A container runs an AI coding assistant on the command line, and that assistant signs in with an AI account you link: Claude Code, Codex, or Gemini. The AI account card on the launch screen lists the accounts you can use, your own linked accounts first and then your organization's, so choose the one the container's AI should sign in with.
- No account linked yet. The card reads Connect a Claude Code, Codex, or Gemini account to use AI in this container. Select Link an AI account and sign in to it; Spin up container becomes available once an account is linked. Link another account adds more later.
- Your organization requires one account. If your organization's policy names the account AI containers must use, the card shows only that account and says your organization requires it. If that account is no longer active, the card says so and you cannot launch until an administrator updates the policy.
The assistant's work runs on the account you linked, not on usage credits. Your paired coding-agent access is restored from your selected Backbuild Secrets vault, and credential files are never saved in the home folder, so a credential does not linger in stored files after the container tears down. See Managing Sessions, Persistence, and Credits for what is restored on a relaunch.
Spin It Up and Watch It Come Alive
After this section you will be able to read every status the launch goes through. Click Spin up container. The terminal attaches immediately and waits, and the container's row in Your containers at the top of the screen shows a status badge for where it is in its lifecycle. You are meant to watch it come up, not guess.
| Status | What is happening |
|---|---|
| Queued | Your request is accepted and waiting for a slot. |
| Scheduling | A machine is being chosen for your container on the backend you picked. |
| Starting | The container is booting and connecting its terminal. |
| Running | The container is live. The terminal is interactive and the editor and desktop tabs are available. |
| Idle | The container is up and connectable but has no work running at the moment. It still draws credits. |
| Closing | The container is shutting down. |
| Terminated or failed | The session has ended, cleanly or because something went wrong, and leaves your containers list. The launch card returns, with Start over, so you can launch again. |
A standard container often goes from Queued to Running in about half a minute; larger sizes and busy periods can take longer. If provisioning takes longer than usual, the screen shows a gentle note that it is taking longer than expected and keeps waiting; the terminal stays attached, so you do not lose your place. When the badge reaches Running, you are ready to work.
What is the cold-start time? A standard container is often usable about half a minute after you click spin up. You watch it progress through Queued, Scheduling, and Starting, and the terminal is attached the whole time.
It has been stuck a while. Did it fail? Not necessarily. A slow provision shows a note that it is taking longer than expected and keeps waiting, with the terminal still attached. A launch that has really failed ends the session: the launch card comes back with Start over. If a launch stays stuck, select Terminate on its row; it works on a container at any stage.
Where to Next
- Terminal, VS Code, and Desktop: connect to and work in your running container.
- Managing Sessions, Persistence, and Credits: switch between containers, terminate cleanly, and understand what persists and what it costs.
- Container Policies and Administration: for admins, what sizes and backends are offered and why.