Autohand Code CLI
Open-source coding agent for the terminal that works with many model providers or local models and asks before taking risky steps.
What it does
Autohand Code works in your terminal: it reads the project, plans changes, edits files, runs commands and handles Git steps such as branches, commits and pull requests. A plan mode lets it examine the code read-only before changing anything. Permission levels and allow lists decide what may run without asking. You choose from more than 15 model providers – such as OpenAI, Anthropic, AWS Bedrock, Azure – or local models via Ollama, llama.cpp and MLX. Skills, MCP servers, a VS Code extension, Zed and JetBrains integration and an SDK for building your own agents round it off.
Who it suits
Development teams that want a coding agent in the terminal or in CI pipelines without committing to one model provider. If you switch from Claude Code, you can import settings and skills.
Data protection for businesses
According to the vendor, Autohand does not train models on your code; if you use your own API keys, the terms of the respective model provider apply. Telemetry and session sync are off by default; the version check (with a device ID) and error reports are on by default and can be switched off in the configuration. The vendor provides a GDPR data processing agreement; processing takes place in New Zealand and the US among other countries, with EU Standard Contractual Clauses for transfers. Only the Enterprise plan lets you control where data lives, and the vendor names no EU region. For companies there are the Team and Enterprise plans, the latter also on-premise.
Limits
The code is licensed under Apache-2.0 with an extra condition: companies with more than US$5 million in annual recurring revenue need a commercial licence. SSO is only on the roadmap. The free plan has tight usage quotas.