An AI coding agent with a persistent workspace

• Files and installed tools

  Files persist across restarts and across chats. Attach the same container to several chats and they share one filesystem. You see the file tree in the web app and can browse, edit, upload and download single files or whole folders as an archive. Edits you make are what the agent sees, with no sync step.

• Pause when idle

  Pausing frees the running slot and keeps the disk until you delete it. At the end of a paid period containers pause and nothing is deleted.

• Snapshots

  Save a known-good state before a risky change. The number depends on the plan: 1 on Mini, 2 on Pro, 3 on Max, 20 on Ultra.

• Export

  Every workspace exports as a Dockerfile and compose file, so the setup is not locked in.

• /goal

  Type what done means, in plain language. After each answer a small model checks the transcript and judges the goal met, not met or impossible. The run ends when the goal is met, stalls after four rounds naming the same missing piece, or hits 25 rounds. It continues if you close the tab. Time per run: 4 hours on Mini, 8 on Pro, 12 on Max, no limit on Ultra.

• Jobs and logs

  Every long-running piece of work gets an entry under Jobs with its state and its log. Follow it live, or read it afterwards. Cost is attributed per job.

• Scheduled tasks

  Give an agent a job and a time: a nightly build, a daily check, a weekly summary. Schedules follow your timezone. Each run is capped, so a loop cannot quietly drain credit, and each run leaves a log.

• You can step in

  A terminal opens in the same container while the agent works. htop, vim and tmux behave normally. Fix one thing by hand, then hand it back. The conversation keeps its context.

ZeroLog Code is a terminal client based on OpenCode. It syncs your local files with the workspace and mirrors console activity and command output into the web chat in real time. The Android app shows chats, jobs and container files on your phone.

Details: Containers, Files, /goal, Jobs and logs, Scheduled tasks, Terminal.

1. Create a container and attach it.

   Pick an image, then attach the container to a chat.

2. Save a snapshot.

   Do it before anything risky.

3. Name a goal you can check.

   Use a condition the checker can see, such as a test run reporting 0 failed.

   /goal pytest reports 0 failed

   Copy
4. Close the laptop. Read the log later.

   The run shows up under Jobs with its state, log and cost.

• Does the workspace survive between chats?

  Yes. The container is durable and a chat is not. Several chats can share one container.

• How many snapshots do I get?

  1 on Mini, 2 on Pro, 3 on Max and 20 on Ultra. Free has no container and no snapshots.

• How long can an agent run on its own?

  One /goal run lasts up to 4 hours on Mini, 8 on Pro, 12 on Max, and without a time limit on Ultra. Rounds are capped at 25 by default, and a run stops when it stalls.

• Can I schedule recurring work?

  Yes: a nightly build, a daily check, a weekly summary. Each run is capped and leaves a log.

• Swiss AI workspace

  Chat with many models and a Linux container in one account.

• Use your own Claude or ChatGPT subscription

  Claude Code and Codex on the plan you already pay for.

• Cloud development environment in Switzerland

  Where your files, your container and the model calls live.

• Claude Code in the cloud

  Your own Claude account, signed in inside a container.

• Codex in a persistent container

  OpenAI's Codex CLI with your own account.

• Open-source models for your coding agent

  The catalogue, how credit pays for it, and your own key.

• ZeroLog Code

  Your workspace, in your terminal.

• ZeroLog Mobile

  Your chats, jobs and container files on Android.