# Agents That Work While You Sleep

Set an agent on a recurring schedule with a budget limit. Dependency updates, test runs, security scans, performance audits — maintenance that runs itself.

Scheduled agents

## Nobody Wants to Do Maintenance. So Nobody Does It.

Everyone knows these things need to happen regularly. Nobody prioritizes them because there's always a feature to ship.

### What happens instead

So maintenance gets pushed until it turns into a production fire drill.

- **Dependencies go stale**
  ignored
- **Test coverage erodes**
  ignored
- **Performance drifts**
  ignored
- **Security patches pile up**
  ignored

### Better default

Set it once. Scheduled agents flip the model: maintenance just happens every night, every week, or every month — without anyone having to remember.

## How it works

### Set It. Forget It. Review When You're Ready.

1. ### Define the task
   “Update all dependencies to their latest stable versions and run the test suite.” Plain English. Same as any Agen task.
2. ### Set the schedule
   Daily, weekly, monthly, or custom cron. Pick the frequency that matches the task.
3. ### Set a budget limit
   Maximum spend per run. The agent stops when it hits the limit. No runaway costs — even if the task takes longer than expected.
4. ### Agent runs automatically
   At the scheduled time, the agent spins up, clones your repo, does the work, fixes any pipeline failures, and opens a PR.
5. ### You review when you're ready
   A PR appears in your queue with a clean diff and passing CI. Review it over coffee. Merge or close. The next scheduled run happens regardless.

## What to schedule

### Tasks That Should Run Themselves

#### Frequency

- ### Daily
  - Dependency updates
  - Nightly test runs

#### Frequency

- ### Weekly
  - Performance audits
  - Security scans
  - Stale documentation check

#### Frequency

- ### Monthly
  - Major dependency upgrades
  - Code quality sweeps
  - License compliance check

## Budget controls

### You Control How Much Each Run Can Spend

When you schedule an agent, you set a maximum budget per run. This is a hard limit — the agent stops when it's reached, even if the task isn't finished.

### Example budgets

- **Nightly dependency check**: $3/run budget  
  Usually finishes under $1.
- **Weekly security scan**: $10/run budget  
  Occasionally needs more if patches are complex.
- **Monthly major upgrades**: $25/run budget  
  Larger scope, higher ceiling.

## How competitors handle scheduling

| Tool                | Scheduled agents?                         | Budget limits?               |
| ------------------- | ----------------------------------------- | ---------------------------- |
| Agen                | ✓ — Any frequency, any task            | ✓ — Hard cap per run       |
| Devin               | ✓ — Recently launched                   | ✗ — No budget controls      |
| Cursor              | ✗                                       | ✗                           |
| GitHub Copilot      | ✗                                       | ✗                           |
| Claude Code         | ✗                                       | ✗                           |

Devin launched scheduling recently — but without budget limits, a scheduled task that runs weekly can accumulate unpredictable costs. Agen's budget controls mean you know the maximum cost before the first run.

## Scheduled Maintenance Across Your Entire Stack

Scheduled agents work with multi-repo sessions. That means you can schedule a task like:

### Example schedule
“Every Monday at 6 AM, update dependencies across frontend, backend, and infra repos. Run all test suites. Open linked PRs if everything passes.”

### Why it matters
No other tool can do this.

## Maintenance on Autopilot

Set the schedule. Set the budget. The agent handles the rest.