codingmoh/open-codex

Fully open-source command-line AI assistant inspired by OpenAI Codex, supporting local language models.

What it solves

Open Codex removes the need to remember or craft shell commands by letting you describe what you want in plain English. It runs entirely on your machine, so no data leaves your computer and no OpenAI API key is required. It also avoids accidental damage by asking for confirmation before executing anything.

How it works

You type a prompt like "find all JPEGs larger than 10MB" and Open Codex sends it to a local language model (either a built-in model like phi-4-mini or one hosted via Ollama). The model returns a suggested shell command, such as find . -name "*.jpg" -size +10M. Open Codex then shows you the command and gives three choices: execute it, copy it to the clipboard, or abort. Only after your explicit confirmation does the command run.

Who it's for

Anyone who works in a terminal and wants a faster, safer way to run commands without memorizing syntax. It is especially useful for developers, system administrators, and power users who prioritize privacy and want to avoid cloud-based AI tools.

Highlights

  • 100% local execution – works offline, no cloud dependency
  • Supports one-shot mode: open-codex "list all folders" returns a shell command
  • Full Ollama integration, letting you use models like LLaMA3 or Mistral
  • Cross-platform: works on macOS, Linux, and Windows
  • Confirmation before running any command, plus copy/abort options
  • Colored terminal output for readability
  • MIT licensed and open source

Related

  • Project
    wlzh/skills

    An open-source collection of automation tools (the 'skills') for Claude Code, covering podcast creation, video processing, text-to-speech, content publishing, code review, invoice scanning, social media fetching, and server security — each usable independently or in combination.

  • Project
    Ar9av/PaperOrchestra

    A pluggable skill pack for coding agents that implements a multi-agent pipeline to transform unstructured research logs and ideas into professional LaTeX research papers.

  • Project
    buzzer-re/Rikugan

    Rikugan is an AI‑driven plug‑in for IDA Pro and Binary Ninja that lets you chat with a large‑language model to run 60+ reverse‑engineering tools, generate patches, and deobfuscate code—all from within the disassembler UI.

  • Project
    bbylw/rea-cn

    REA is a Node‑JS platform that connects LLM agents to reverse‑engineering tools (Hopper, Ghidra, IDA, etc.) and static analysis utilities for binaries, JavaScript/Electron, .NET, Android, firmware, web captures and more. It runs a local MCP service, registers a skill with supported agents, and returns concrete evidence (pseudo‑code, assembly, module graphs, screenshots) that the AI can use to explain or re‑implement a target feature. Install via `npx rea‑agents setup`, then ask questions or use the CLI (`rea …`).