We recently open-sourced Rover, a manager for AI coding agents. It is a tool we developed internally to help us work better with coding agents like Claude and Gemini.

If you take a look at some of HN stories or tweets, it seems many developers are getting amazing results from coding agents while the rest of people are completely disappointed to the point of being a net negative.

Think of Rover as the bundling of best practices and technical functionality (i.e. git worktrees) to have an easy way to interact better with coding agents. You can run multiple agents in the background and even combine them to complete a task.

In practice, what this means is that:

You can assign tasks to multiple agents and they can work independently and in parallel. The agents don’t interfere with each other or your system. They run inside containers and use git worktrees. Everything is transparent and runs locally. It installs as an npm package in the command line or as an extension to VSCode. You do not need to change anything in the way you work. We now support Claude, Codex, Gemini and Qwen.

Give it a try and let us know what you think!

https://github.com/endorhq/rover