I built Group Leveling, a self-hosted collaboration stack for small groups of friends who want a shared workspace for human chat, local repositories, and coding-agent workflows.
Repo:
https://github.com/leonidas1312/group-leveling
The idea is simple: one person hosts it on an old laptop or small home server, invites friends over LAN or Tailscale, and everyone gets a shared workspace for chat, repos, and agent-driven coding work.
The stack includes:
- Next.js app for the group workspace
- Gitea for local repositories, branches, and pull requests
- A local Codex workflow server
- Optional Tailscale mode for private access
The app supports shared chats where people can mention teammates with @username, mention agents with @agent-name, and reference repositories with #owner/repo.
Example:
@agent-alex #alex/app fix the settings bug
That starts a workflow against the local Gitea repo. The agent belongs to a user, and the workflow runs through that user’s connected ChatGPT/Codex auth.
Why I built it:
I wanted something my friends and I could host ourselves instead of relying on a hosted team-agent product. The goal is to make a small-group coding workspace that is easy to run, customizable, and private by default.
I’d love feedback on:
- whether the self-hosting setup is clear enough
- what security/privacy concerns would stop you from running this
Top comments (0)