Skip to main content

​
Why

Combine AI with automation. Instead of pre-written notebook cells, give an AI agent a prompt and a schedule. The agent decides what to do each run — adapting to new data, changing conditions, or evolving requirements.

​
Setup

1

Go to Automations

In your project, click Automations → New Automation
2

Choose Agent

Select Agent as the automation type
3

Configure

  • Write a prompt describing what the agent should do
  • Select an AI model (Claude, GPT, or Gemini)
  • Set a name and cron schedule
  • Click Create

​
Example Prompts

Daily data summary:
Analyze today’s sales data in sales_2024.csv. Write a summary to daily_reports/ with key metrics and any anomalies.
Code health check:
Run the test suite. If any tests fail, investigate the failures and create a report in debug_logs/.
Research digest:
Check the latest papers on arxiv related to diffusion models. Summarize the top 3 in a new markdown file.

​
How It Works

The agent gets your prompt, access to the project’s filesystem and kernels, and runs autonomously. It can create files, execute code, read outputs, and iterate — just like an interactive IDE session, but unattended.

​
Managing

  • Toggle automations on/off
  • View run history to see what the agent did, including generated files and outputs
  • Edit the prompt, model, or schedule anytime
Agent automations use AI tokens billed to your organization. The project must be running for automations to execute. Default timeout is 300 seconds.