About

Built for the developer without their own NOC

Tailon is a small, independently built product, not a team with its own support department. Just someone who ran into the same noise you did: every Sentry notification landing directly on you, with no way to tell if it was something or nothing.

Tailon is what was left after cutting away the rest. One sharp thing, done well.

The story

Tailon started broader: a hosting platform for AI agents, inspired by myclaw.ai. After a few days of building, that setup turned out bigger than the actual problem. The real problem was somewhere else: too many Sentry notifications, too little time to read all of them. Tailon is what remained after cutting away the rest. One sharp thing, done well.

The philosophy

Three rules sit above everything else: when in doubt, Tailon always rounds urgency up, never down. Only a fixed whitelist of fields ever leaves your Sentry environment, never the full payload. And every diagnosis is a hypothesis, not a guarantee: Tailon never takes action on its own, it only suggests what you should look at.

Built in public

Every decision, misstep, and course correction is documented in the public changelog, at a milestone level, not per commit. No polished press release after the fact, just the notes as they happened, including the things that didn't go right at first.

What's still missing

Tailon is a working core, not a finished end product. Still open: a balance check before a triage runs, a screen where you pick your own AI model, and connections to more sources than Sentry alone. It's listed here so you know where things stand now, not only once you run into it yourself.

Roadmap

What's built, what's next

1Now
Core product
  • One-webhook Sentry connection
  • AI triage with a fast + a thorough model
  • Email only for urgent incidents
  • One subscription, no credit bookkeeping
2Next
More sources, more channels
  • Datadog and PagerDuty as extra sources
  • Slack, Discord and Telegram as channels
3Later
Pick your own model
  • Choose from 200+ models via OpenRouter
  • Not just fast-vs-thorough anymore