What Opsh does
When you open a terminal with Opsh installed, it starts automatically and gives you an interactive prompt. You type a task in natural language — like “find all log files older than 7 days” — and Opsh sends that request to your configured AI provider. The provider returns a shell command. Opsh shows you the command, a brief explanation, and a risk classification, then waits for you to confirm before running anything. You stay in control at every step. Opsh never runs a command without showing it to you first (unless you enable warp mode, which auto-runs commands classified as safe).Key benefits
- No syntax memorization — describe tasks in plain English instead of looking up flags and options
- Transparent execution — every generated command is shown with an explanation and risk level before it runs
- Multiple AI providers — choose OpenAI, Anthropic, Gemini, OpenRouter, or run a local model via Ollama
- Shell history context — Opsh reads your recent command history to generate more relevant suggestions
- Warp mode — automatically runs safe commands and summarizes the output in plain English
- Print-only mode — preview generated commands without executing them, useful for scripting and auditing
Supported platforms and shells
Opsh works with bash and zsh. It detects your shell automatically on startup, and you can override this with
--shell zsh or --shell bash at any time.
Installation
Install Opsh on macOS or Linux with a single curl command.
Quick Start
Set up your AI provider and run your first natural-language command in minutes.
Configuration
Configure your AI provider, model, shell preference, and execution behavior.
Warp Mode
Let Opsh auto-run safe commands and summarize results in plain English.
Opsh requires an API key from a supported AI provider — OpenAI, Anthropic, Gemini, or OpenRouter. If you prefer to keep everything local, Opsh also supports Ollama, which runs models on your own machine with no API key required.

