Helix Agent

The Helix Agent is the autonomous worker that reads your instructions, explores repositories, runs tests, and interacts with connected cloud providers.

Modes

• Chat: Ask questions about your codebase or cloud setup.
• Task: Assign a specific outcome and Helix will plan and execute it.
• Operator: Helix runs continuously and asks for approval before destructive actions.

Autonomy levels

Each plan has a configurable autonomy level. Read-only work is always allowed. State-changing actions require explicit approval by default. Enterprise teams can add additional approvers and scheduled windows.

Conversation context

Helix keeps a running context of the conversation, repository state, and tool outputs. Long outputs are summarized to stay within context budgets. You can pause, resume, or cancel a run at any time.