An agent is the coding assistant Helmor orchestrates during a session. Helmor handles workspace isolation, task context, and review; the agent performs the actual coding work through its provider CLI.

What agents do inside Helmor#

During a session, an agent can:

• Read project files — understand the codebase, imports, and structure.
• Edit files — write new code, refactor, or fix bugs in the workspace.
• Run commands — execute shell commands (build, test, lint) when the environment permits.
• Report progress — stream status, ask clarifying questions, and surface verification results.

All agent activity is visible in the session timeline, so you can follow along in real time.

Supported agents#

Helmor supports CLI-backed coding assistants. Current options include:

Each provider manages its own authentication independently of Helmor. Make sure the agent CLI is installed, on your PATH, and authenticated before starting a session.

Choosing the right agent#

Pick the agent that matches your task:

• Implementation & refactoring — both Claude Code and Codex handle multi-file edits well.
• Debugging — agents with strong reasoning (Claude Code) excel at tracing issues.
• Code review tasks — use whichever agent you trust for stylistic and correctness feedback.

Your responsibility as the operator#

Agent output is a draft, not a finished product. Always:

1. Review the diff before committing.
2. Run tests and checks.
3. Decide what ships — Helmor makes the review loop visible but does not remove the need for engineering judgment.

Related#

• Changes and Review — how to inspect and ship agent work.
• Sidecar and CLI Issues — troubleshoot agent startup failures.