Watch a small fleet of self hosted servers for downtime, drift, and security issues from one dashboard
Get a plain language explanation from Claude when disk space or memory usage looks abnormal
Deploy code with a typed confirmation gate plus automatic post deploy verification and rollback
Receive a Telegram page only for genuinely critical incidents instead of constant noise
| dcblack87/atlas | 0xustaz/streamgate | a-bissell/unleash-lite | |
|---|---|---|---|
| Stars | 1 | 1 | 1 |
| Language | Python | Python | Python |
| Setup difficulty | moderate | hard | hard |
| Complexity | — | 4/5 | 4/5 |
| Audience | ops devops | developer | researcher |
Figures from each repo's GitHub metadata at analysis time.
Needs SSH access to your servers over a tailnet plus a Claude API key for AI explanations, demo mode works without any of that.
Atlas is a terminal based operations tool built for a solo developer running their own servers, instead of a whole team with a big monitoring budget. It runs continuously in tmux on one of your own machines and connects to Claude for explaining problems in plain language. Once set up, Atlas automatically finds the hosts, containers, websites, certificates, cron jobs, and backups on your fleet by connecting over SSH, with nothing needing to be installed on the machines it watches. It shows this on a dashboard answering three basic questions: is anything down right now, is anything about to break, and is there anything ready to deploy. When your code is ahead of what is actually running on the server, that shows up as a tile you can act on with one confirmation. When something goes wrong, Atlas does not just report a raw number like disk usage percentage. It uses Claude to explain what happened in context, for example noting that disk space grew because of build cache from recent deploys and suggesting the command to clear it. AI usage is capped by a hard daily budget enforced in code, and spend is tracked against that budget. Atlas also learns your server's normal patterns over a few weeks so it can flag things that look unusual and forecast issues like a disk filling up before it happens. A notable design choice is that Atlas is read only by default. Only one specific part of the codebase is allowed to run commands that change anything on your servers, and every one of those changes requires a typed confirmation, is limited to an approved list of actions, and is logged in an audit trail. It never needs cloud account credentials beyond a read only billing token. The project is also built to run well on low power e-ink tablets and displays alerts through Telegram for critical issues only. It requires Python 3.12 or newer and is licensed under MIT.
A one person operations dashboard that watches your own servers over SSH and uses Claude to explain problems and flag risky deploys.
Mainly Python. The stack also includes Python, SQLite, SSH.
Use freely for any purpose, including commercial use, as long as you keep the copyright notice.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly ops devops.
This repo across BitVibe Labs
double-check against the repo, no cap.