DEV Community

Cover image for I built a CLI that lets me use AI directly from my terminal
Alexandre Vieira
Alexandre Vieira

Posted on

I built a CLI that lets me use AI directly from my terminal

I’ve been building Xandai-CLI because I wanted AI help without leaving the terminal or wiring up cloud APIs.

The idea is simple: one CLI where I can run shell commands, edit files, and ask questions in the same place — using local models (Ollama / LM Studio).

Repo: https://github.com/XandAI-project/Xandai-CLI

What Xandai-CLI does

Xandai-CLI is an interactive terminal tool that mixes:

normal shell commands

natural language prompts

file creation and editing

optional multi-step “agent” behavior

You type at the xandai> prompt, and it figures out whether you meant a command, a question, or a task.

Examples:

xandai> git status
xandai> explain what this repo does
xandai> create a python script to parse this log file

Everything runs locally if you want it to.

Why I built it

Most AI coding tools either:

live in the browser, or

require API keys and cloud calls, or

feel bolted on to the terminal

I wanted something that felt closer to git or fzf:
always there, fast, and not intrusive.

Xandai-CLI is opinionated around developer workflows and local control.

Local models first

Xandai-CLI supports:

Ollama

LM Studio

You can auto-detect what’s running and start using it immediately. No accounts, no tokens.

Install & try it
pip install xandai-cli
xandai --auto-detect

From there, just use it like a normal shell — but with the option to ask for help or generate code when you need it.

Status

This is still evolving, and I’m actively iterating on it based on how I actually use it day-to-day.
If you try it and something feels off, or you have ideas, issues and PRs are welcome.

Top comments (0)