When the AI intern gets direct write access to your repo

What it does

o1-engineer acts as a thin orchestration layer between your terminal and remote LLM APIs. You issue slash commands to load files into context, generate code, plan project structures, and apply edits directly to your local filesystem.

The interesting bit

The tool treats your repository as mutable chat state. Instead of copy-pasting snippets, you issue commands and the AI edits files in place. The /planning command is a neat addition: outline an entire project structure in conversation, then materialize it with /create.

Key highlights

• Supports both OpenAI and xAI (Grok) APIs, with streaming responses recently added.
• Slash-command interface: /add, /edit, /create, /review, /planning, /reset, and /debug.
• Operates on both individual files and entire folders for batch context loading or modification.
• Maintains conversation history across the session and allows clearing context to start fresh.

Caveats

• OpenAI API keys are configured by pasting them directly into the script source, while the newer xAI integration expects a .env file—an inconsistent setup.
• The README never specifies which models are targeted (despite the “o1” branding) or how it handles large file contexts.
• AI-generated changes are applied directly to disk; the README does not mention any confirmation step, diff preview, or backup mechanism.

Verdict

Try it if you want a lightweight, chat-driven way to scaffold or refactor code without switching windows. Avoid it if you need offline access, deterministic builds, or guardrails before an LLM mutates your working tree.