Ona Agents extend your engineering team’s capacity by completing tasks securely in isolated environments. Agents can propose and apply changes across repositories with full environment context, while respecting your organization’s guardrails and policies.

​
What agents can do

  • Generate or modify code, tests, and documentation
  • Run commands, formatters, and linters inside isolated environments
  • Open branches and pull requests, and respond to review feedback
  • Execute long-running tasks (e.g., refactors, migrations, test runs)
  • Follow organizational controls such as command allow/deny lists and SSO

​
Modes of operation

  • Manual: You drive; agents assist with suggestions or targeted actions.
  • Assisted: Stay in control while agents propose and apply changes with full visibility.
  • Autonomous: Delegate long‑running or repetitive work to agents running in isolated environments.

​
Prerequisites

  • An Ona runner and environment setup (Ona Cloud or your AWS account)
  • Dev Container and Automations configured for your repository
  • LLM provider configured (enterprise only) — see LLM Providers

​
Quick start

  1. Configure environments and runners:
  2. Bring your LLM provider (enterprise only):
  3. Add team conventions and workflows:

​
Best practices

  • Keep environment definitions reproducible with Dev Containers and Automations
  • Start in Assisted mode, then graduate tasks to Autonomous as confidence grows
  • Use guardrails (policies, command deny list) for safe, auditable execution
  • Test agent workflows in non‑production repositories before broad rollout
Next steps: