Maximize Productivity with Manager Surface
Antigravity's Manager Surface is a powerful feature that lets you run multiple AI agents simultaneously for parallel task processing. However, simply spinning up agents won't unlock its full potential. Here are five techniques that make a real difference in practice.
1. Choose the Right Task Granularity
The tasks you assign to agents should be neither too broad nor too narrow.
Good examples:
- "Refactor the validation logic in the UserService class"
- "Add error handling tests for the API responses"
Avoid:
- "Refactor the entire app" (too broad)
- "Rename variable x to count" (too narrow)
As a rule of thumb, each agent should be able to complete its task within 15 to 30 minutes.
2. Explicitly Specify Context Files
Agents are smart, but explicitly pointing them to relevant files dramatically improves accuracy.
@agent frontend-fix
context: src/components/UserForm.tsx, src/hooks/useValidation.ts
task: Change form validation to real-time
3. Set Dependencies Between Tasks
Manager Surface allows you to configure dependencies between agents. For example, you can chain "Update API type definitions → Sync frontend types → Update tests," ensuring each agent references the output of the previous one.
4. Set Review Checkpoints
When running three or more agents simultaneously, don't wait until everything is done. Set intermediate checkpoints. The Manager Surface dashboard shows real-time progress for each agent.
If you notice an agent drifting off course, correct it early to minimize rework.
5. Use Templates for Recurring Workflows
Save frequently used agent configurations as templates. For instance, a "New Feature Pack" template might include:
- Agent 1: Feature implementation
- Agent 2: Unit test creation
- Agent 3: Documentation updates
Conclusion
Manager Surface becomes exponentially more powerful as you master it. Start with two agents, and gradually increase parallelism as you get comfortable.
Stay tuned for more practical techniques from Antigravity Lab.