Skip to main content
When your agent is defined as markdown files for a coding agent (Claude Code, Codex, Cursor, OpenCode): its system prompt, config, and instructions as files in your repo, Helix applies approved fixes by editing those files and opening a pull request on GitHub. Supported today: Claude Code, Codex, Cursor, OpenCode, and skill directories. Pi and Hermes are coming soon.

Claude Code

Agents defined in .claude/agents/*.md.

Codex

Agents defined as Codex markdown / TOML.

Cursor

Agents defined as Cursor markdown.

OpenCode

Agents defined as OpenCode markdown.

Skill directories

A skill folder (SKILL.md with its references and assets). Set root to the skill’s folder.

Pi (coming soon)

Agents defined for the Pi runtime.

Hermes (coming soon)

Agents defined for the Hermes runtime.

​
How a fix lands

Helix never writes to your repo directly. After Diagnose ranks a remedy and you approve it:
  1. Helix prepares the change on a branch.
  2. It opens a GitHub pull request with the diff.
  3. You review and merge — the change is yours to accept.
Every apply is approval-gated. Helix proposes; GitHub is where you accept. Nothing lands on your default branch without your merge.
Before you start: the repository has a GitHub remote, and gh auth status succeeds or you can push over SSH. Helix opens the pull request with those credentials. Declare a markdown agent target in your config:
.mutagent/config.yaml
Run helix -p "/status" (it needs a working model key) to check that the target is set up. Building an agent as code instead of markdown? See code targets.