A side project that runs while you sleep

Updated 1 October 2026 · BugsRadar team

It started as a weekend project and now it has users. It also has no monitoring, because monitoring is a chore and the project is supposed to be fun. So the first alert is a user writing "is it down for you too?", three days after the fact.

Ten minutes, once

Sign in at app.bugsradar.com with your email, create a project, and add a Telegram channel: a bot from @BotFather and a private chat with it, no group needed. Add the package for .NET, Node.js or Python with the key in an environment variable, and everything the app logs at Error and above, plus every uncaught exception, now arrives in that chat. Send a test error, see it land, close the laptop.

Nothing to pay, nothing to babysit

The Free plan is a plan, not a trial: 2 projects, 3 channels, no event quotas, no card. There is no dashboard to check and nothing to clean up, because BugsRadar stores nothing; the message in your chat is the whole record. If the project turns into something bigger, Pro is $5 a month: see Pricing.

What a bad night looks like

At 02:14 the database connection starts failing. You get one message with the exception, where it happened and the stack trace. By morning the same message says ×412 and nothing else has arrived, because repeats are counted in the first message instead of sent again. You fix it over coffee. If you would rather not be woken at all, mute the chat for the night; it is your chat and your bot.

The VPS it runs on

Side projects live on a small server with a cron job and a backup script that nobody has checked since the day they were written. Both can report their own failure with one HTTP request into the same chat: see the guides for cron and backups.

What you don't get

History. There is no page of last week's errors; if you want to look back, keep the app's logs. For a side project that is usually the right trade: you want to know the moment it breaks, not maintain an archive of the times it did.

Where to go next

Get started is the ten minutes. Silent failures is why the first alert should not be a user.