Most of my vibe coding turned into a mess of folders, terminals and half finished apps I couldn't get running again a month later. I run a small computer repair shop and build apps on the side, so I don't have a lot of time to spend remembering how each one starts. So I made Zeltro to keep it all in one place.
You type what you want to build, it suggests a setup and tells you why, then your AI gets to work in a terminal right inside the app. When the AI needs an API key it asks for it in a private box, so the key never ends up in the chat. You get a little pop up when it's done or needs you, so you can go do something else in the meantime.
Every project gets its own address so nothing fights over ports, and they all share one database so you're not setting up a new one every single time.
It works with Claude Code, Codex, Gemini, Qwen, or a local model. The desktop app is free for personal use, and the command line tool underneath is open source.
It doesn't sell you any AI either. You use the key or plan you already pay for, or a free local model, and pay the provider's own price instead of buying credits. It's still in beta, so I'm looking for people to break it. I'd love to hear what you think in the comments.
You can grab it at zeltro.ai.


Top comments (1)
This is a really practical idea! I like how Zeltro addresses a problem that many developers face with AI-assisted coding: managing multiple projects, keeping environments reproducible, and getting back to work without having to reconstruct everything from scratch. The secure handling of API keys and support for different AI coding tools are especially interesting.
I’m currently exploring AI agent development with LangChain, focusing on agent workflows, tool integration, and reliable execution. I see some interesting possibilities for combining project management with AI agents that can help developers initialize projects, run tests, diagnose errors, and maintain useful context between coding sessions.
I’d love to learn more about how you designed Zeltro’s architecture, particularly how you manage isolated project environments and integrate different coding agents.
Have you considered adding support for custom agent workflows or LangChain-powered automation? I’d be interested in exchanging ideas and potentially collaborating on a small proof of concept that connects AI agent workflows with a project management environment like Zeltro.
Great work turning a real everyday frustration into a practical developer tool. I’m looking forward to seeing how the project evolves!