01
Parallel execution
Run and test independently.
One agent starts an app. Another tests a change. A third uses the database. Give each its own environment so their work can happen at the same time.
Execution environments for AI agents
Run your agents.
Side by side.
Give each agent its own environment to run apps, use services, and test changes—without colliding with the others.
On your machine. In your VM. On your compute.
One project. Three execution tabs.Example
Checkout environmentIndependent
- App
- localhost:3000
- Database
- checkout_db
- Browser
- checkout_session
Same app port. Separate databases and browser state.
Code can be separate.
Execution should be, too.
Worktrees separate the files. Tabs gives each agent its own place to run them, with independent apps, services, and test environments.
Less environment work. More useful agent work.
Three things your agents need.
02
Ready to run
Prepare once. Start from there.
Start with a ready-to-run agent image. Reuse prepared environments instead of rebuilding the same setup for every task. Add your project's dependencies where needed.
03
Least privilege
Access for the task at hand.
Choose the files and network access each task needs. Let the agent work within those boundaries, without opening up the rest of your environment.
A tab for every task.
A consistent way to start work, inspect it, and clean up.
- 01
Open
Choose a prepared environment.
- 02
Run
Set permissions. Start the work.
- 03
Inspect
Follow progress and collect results.
- 04
Close
Dispose of the task environment.
Tabs / Early access
What would your
agents run?
Tell us about your workflow. We're building Tabs for developers and teams running agents on their own compute.