Turn a plain English request into a shell command without memorizing syntax
Review, edit, or reject an AI-generated command before it runs on your machine
Connect the tool to any local LLM server like llama.cpp or Ollama
Keep a running memory of past requests and commands within a project folder
| vkataev/ai-cli | angree/openttd_amiga_68k | drakeee/samp-plugin-lua | |
|---|---|---|---|
| Stars | 4 | 4 | 4 |
| Language | C | C | C |
| Last pushed | — | — | 2015-07-22 |
| Maintenance | — | — | Dormant |
| Setup difficulty | moderate | hard | hard |
| Complexity | 3/5 | 4/5 | 3/5 |
| Audience | developer | general | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires compiling with gcc and a separately running local LLM server to connect to.
This is a small command-line tool that lets you type a request in plain English and get it turned into a shell command that actually runs on your computer. Instead of remembering the exact syntax for things like finding files, changing permissions, or searching text, you type something like ai find all lines containing perform in this file, and the tool sends that request to a local AI language model, gets back a shell command, and shows it to you before running anything. You stay in control the whole time. When the suggested command appears, you can press Enter to run it as is, edit it first using arrow keys like a normal text editor, or press Ctrl+C to cancel and do nothing. There is also an optional memory mode that saves a running history of your requests and the commands used, so the assistant can refer back to earlier context in the same folder. The entire tool is written as a single C file with no external dependencies, which the author built specifically so it could be compiled and run on almost any system, including Linux, macOS, Windows, and several less common operating systems. It does not include its own AI model. Instead, you point it at any local language model server you already have running, such as llama.cpp or Ollama, by setting an environment variable to that server's address. The project was made by someone who wanted to stop switching back and forth between a terminal and a separate chat window just to ask how to write a one-line command. Because the tool executes whatever the AI model returns directly in your shell, the author is explicit that this carries real risk: if you are not comfortable reading and understanding shell commands before running them, this tool is not meant for you.
A tiny command-line tool that turns plain English requests into real shell commands using a local AI model, letting you review or edit them before they run.
Mainly C. The stack also includes C.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
double-check against the repo, no cap.