FAQ
Frequently asked questions about Hanzo Dev.
Which models are supported?
We recommend the built-in presets that wrap GPT-5.2 (e.g., code-gpt-5.2-codex). The default reasoning level is medium — upgrade to high for complex tasks with /model.
You can also use other models via API-based auth with the --model flag.
How do I stop Hanzo Dev from editing my files?
By default, Hanzo Dev can modify files in your current working directory (Auto mode). To prevent edits, run in read-only mode:
dev --sandbox read-onlyOr change the approval level mid-conversation with /approvals.
Does it work on Windows?
Running directly on Windows may work but is not officially supported. We recommend Windows Subsystem for Linux (WSL2).
Why can't Hanzo Dev find my agents on Windows?
Agent discovery on Windows can be affected by PATH and file extensions. Solutions:
Use absolute paths (recommended)
[[agents]]
name = "claude"
command = "C:\\Users\\YourUser\\AppData\\Roaming\\npm\\claude.cmd"
enabled = trueFind your npm global install location
npm config get prefixVerify PATH includes npm directory
$env:PATH -split ';' | Select-String "npm"Check file extensions
On Windows, Hanzo Dev looks for executables with extensions: .exe, .cmd, .bat, .com. Include the correct extension when using absolute paths.
What is the relationship to OpenAI Codex?
Hanzo Dev is a Rust-native agentic development CLI: multi-agent orchestration, MCP client and server, a sandboxed execution policy, and direct integration with Hanzo AI infrastructure. The name is unrelated to OpenAI's 2021 Codex model, which was deprecated in March 2023.
How is this guide?