devspace

DevSpace lets you use ChatGPT as a real coding agent that works directly with your local files. Instead of copying and pasting code into a chat window or uploading your project somewhere, you give ChatGPT a secure, controlled connection to specific folders on your own computer. From there, it can read your code, make edits, search through directories, run shell commands like tests and builds, and show you what changed — all while you watch and approve what's happening.

At a high level, it runs as a small server on your machine. You install it through the command line, pick which project folders ChatGPT is allowed to access, and expose it through a tunnel you control (like Cloudflare Tunnel, ngrok, or Tailscale). ChatGPT connects to a public URL you provide, and you approve the connection with a password that only you hold. Once connected, ChatGPT can open a workspace from your approved folders and start working — reading files, editing code, running git commands, and even using isolated git worktrees for parallel sessions. It also reads project instruction files like AGENTS.md and CLAUDE.md so it can follow your project's conventions.

This is built for developers, founders, or anyone who already uses ChatGPT for coding help but is tired of the back-and-forth of pasting snippets manually. If you've ever wished ChatGPT could just look at your actual project, run your test suite, fix what's broken, and report back — this is that. It works on Linux, macOS, and Windows (with a Bash-compatible shell like Git Bash or WSL).

The project is written in TypeScript and runs as a self-hosted server, meaning you stay in control of your data and your machine. Nothing gets uploaded to a third party. The tradeoff is that you need some comfort with the command line and setting up a tunnel, and the README is clear that a connected AI client has real, powerful access to your system — so you should treat it like giving a trusted coworker access to your terminal. There's also an optional sponsor program that can show a small ad in your coding sessions and pay you cash back, which is an unusual monetization choice for an open-source tool.