CoCalc feature
Online Linux Terminal
A Linux terminal for you and your agents.
Agents can run commands in a CoCalc project. Open its Linux terminal to check or rerun them, work with collaborators, and save outputs before the project stops.
Agents run commands and save their work in the project. Open a terminal to check the output, rerun a command, or work directly.
- Agents can run project commands
- Files and results stay in the project
- Open a shell for direct control
- Share a live terminal with a collaborator
Run build tools, Git, and tests
The terminal has the software in your project's image. If your project has network access, add packages with the image's package manager, such as sudo apt-get install.
Bring a collaborator into the same shell
Collaborators on the project can open the same terminal and see the same output live.
Let the agent work in your terminal
When AI is enabled for your site, account, and project, click the Agent button in a terminal's title bar, and the agent can read that session and run commands in it. Commands it runs on its own show in its chat, not in your terminal, so check the files and output, or rerun a command yourself.
With network access, you can also install and run other command-line agents as ordinary Linux tools.
Keep the result after the process ends
A browser disconnect does not itself terminate the backend shell. You can reconnect while the project runtime is still running. Project stops, restarts, failures, or a configured browser-idle timeout end running processes, even during a computation. Save logs and checkpoints so important jobs can resume.