mang.sh

Product Introduction

1. Definition: mang.sh (originally developed as yo-rust) is a command-line interface (CLI) tool written entirely in Rust that translates natural language prompts into precise terminal commands. Categorized under AI-powered developer tools and natural language CLI utilities, it operates via single compiled binary files with zero runtime dependencies. It interfaces with either cloud-based AI providers via OpenRouter (using GPT-4o-mini as a default) or local privacy-first models running through Ollama.
2. Core Value Proposition: mang.sh exists to eliminate the "translation tax" developers pay when converting conceptual intents into exact syntax for POSIX shells, Zsh, Bash, Fish, or PowerShell. By bridging human intent with machine execution via natural language shell commands, the tool provides a safety-first workflow where every generated command requires explicit user confirmation before execution, reducing syntax lookups, preventing destructive errors, and boosting developer productivity.

Main Features

1. Always Confirms Safety Guardrails: Every natural language request processed by mang.sh undergoes an evaluation step where the AI returns a formatted command alongside a one-sentence explanation. Nothing executes automatically. Destructive operations (such as rm, drop, or format commands) default to a "No" response, protecting users from accidental data loss. Users can also leverage dry-run mode (yo --dry) to preview commands without executing them.
2. Dual-Backend AI Support (OpenRouter and Ollama): The CLI offers flexible model routing. Developers can use OpenRouter for cloud-powered models like GPT-4o-mini or switch to local models via Ollama (use ollama or change model) to run workloads offline. This guarantees absolute data privacy by keeping prompts and command generation entirely localized when desired.
3. Multi-Turn Context and Interactive Wizards: mang.sh maintains a full session history, enabling conversational follow-ups such as "now do the same for /tmp" without retyping context. Additionally, the integrated prompt wizard (!prompt) provides a guided three-question clarification mode for vague requests, while named shortcuts (!save <name>) allow users to cache frequent command sequences for instant, AI-free replays.

Problems Solved

1. Pain Point: Developers frequently waste time context-switching between the terminal, search engines, and documentation sites to look up complex CLI flags, regular expressions, or syntax variants for tools like git, Docker, and awk. This manual translation overhead breaks flow state and introduces syntax errors that can corrupt local files or production systems.
2. Target Audience: System administrators, DevOps engineers, software developers, site reliability engineers (SREs), and command-line power users who work extensively in UNIX or Windows shells and want to accelerate repetitive tasks without sacrificing safety.
3. Use Cases: Rapidly querying disk utilization and cleaning up large files, parsing complex system logs and process lists, constructing intricate Git history queries, and automating ad-hoc shell operations across macOS, Linux, and Windows environments.

Unique Advantages

1. Strengths & Limitations (Pros & Cons):
   • Strengths: Exceptional performance with sub-10ms startup times due to its Rust foundation; zero runtime dependencies (no Node.js or Python environments required); cross-platform compatibility (macOS, Linux, Windows); built-in safety confirmations; native integration with shell history (~/.zsh_history or ~/.bash_history).
   • Limitations: Requires an active internet connection and OpenRouter API credits unless configured with a local Ollama instance; AI-generated commands can occasionally hallucinate flags or edge-case arguments, necessitating careful manual inspection despite the built-in confirmation prompts.
2. Key Alternatives & Differentiation: Unlike Python- or Node-based AI CLI wrappers that suffer from heavy interpreter cold starts and dependency hell, mang.sh distributes a single lightweight Rust binary. Compared to cloud-locked AI terminal extensions, mang.sh supports air-gapped local execution via Ollama and enforces strict default-to-safe confirmation protocols rather than blindly executing generated code.

Frequently Asked Questions (FAQ)

1. How do I install mang.sh on macOS or Linux? You can install mang.sh using the official single-line shell installation script by running curl -fsSL https://mang.sh/install | bash in your terminal. For Windows PowerShell users, the equivalent installation command is iwr -useb https://mang.sh/install.ps1 | iex.
2. Can I use mang.sh offline without sending data to the cloud? Yes. By configuring mang.sh to use a local Ollama backend (such as llama3.2), all natural language processing and command generation happen entirely on your local machine with zero network requests or telemetry.
3. Does mang.sh execute terminal commands automatically? No. mang.sh strictly enforces a human-in-the-loop workflow. It generates the terminal command and a brief explanation, but requires explicit user confirmation before running anything, and automatically defaults to "No" for destructive operations.
4. How do I save frequently used commands for quick reuse? You can save any generated command sequence by typing !save <name> during your session. Once saved, you can replay the command instantly by typing !<name> without triggering an AI round-trip.