Hey everyone.
This is my first open source project, and I’m looking for honest feedback rather than trying to promote a finished product.
I built a small self-hosted tool that lets me control Codex CLI or Claude Code from Telegram. The idea is that the bot, server, repos, and agent credentials stay on my own machine/server.
The workflow I wanted:
- send a coding task from Telegram
- run it against a configured local repo
- inspect logs, git status, diffs, and checks
- require manual approval before deploy or rollback
- optionally run multi-agent “missions” with roles like product, designer, QA, frontend, backend, and builder
It is still MVP-grade, and I know there are things to improve:
- better Mini App auth
- real task queue
- more tests
- safer server profile storage
- GitHub PR creation
- better docs and onboarding
I’d appreciate feedback on:
- whether this idea is actually useful
- what looks unsafe
- what architecture choices I should rethink
- what would make it production-ready
GitHub link, if anyone wants to inspect the code: Git
Top comments (0)