When it crashes, your AI reads the logs and fixes it
The hard part of running a bot is not deploying it. It is the morning you find it stopped, with no idea why. Redon3 keeps the evidence and lets ChatGPT or Claude read it, so the fix happens in a chat instead of a debugging session.
Most hosting loses the crash
A container exits and takes its output with it. What you get is a status that says stopped, and no way back to the moment it broke. So you restart it, watch it, and wait for it to happen again.
That is also why asking an assistant for help usually goes nowhere. Without the actual error it is guessing from your description, and you end up pasting fragments between two windows.
The evidence is still there afterwards
- Recent output after the process has exited, not only while it runs.
- The exit code, so a clean stop and a failure are told apart.
- The restart count, which is what a crash loop actually looks like.
- A live log stream while it runs.
- A real terminal in the browser for the times you want to look yourself.
- The files exactly as they were when it died.
- A failing container is restarted up to five times before it is left down.
How a fix actually goes
- You notice, or you ask. Tell your assistant the panel is down. It does not need you to describe the symptom, because it can look.
- It reads the logs. Recent output with the exit code, straight from the panel over MCP.
- It edits the file. A missing dependency, a bad path, an unhandled response, a variable that was never set.
- It restarts and checks. It presses restart, reads the new output, and tells you whether that held.
What it may do while fixing
It may read the logs, overwrite the broken file, change the runtime, start file, or port, add a missing environment variable, and stop, start, or restart the panel.
It may not delete your panel, since there is no tool for it. It cannot buy or renew anything, cannot read back a secret you already stored, and cannot touch another account or anything outside your panel folder. The worst it can do is write a bad file you can overwrite.
Common questions
- Can my AI restart my bot?
- Yes. Start, stop, and restart are three of the 28 tools it gets once you add Redon3 as a connector in ChatGPT or Claude.
- Does the panel restart itself if it crashes?
- Yes, on failure, up to five times. After that it stays down so a broken deploy does not flap forever, and the logs and exit code are waiting for you or your assistant.
- How far back do the logs go?
- Recent output is kept and stays readable after the process exits, which is the part that matters for finding a crash. It is a rolling window, not a permanent archive.
- Will it fix anything on its own without asking me?
- It acts when you ask it to. Nothing here runs an assistant on a schedule or lets it change your panel while you are away.
- What if the crash is not in the logs?
- Then you have a terminal in the browser and the files as they were. Being killed for memory is the common silent case, and the panel shows the RAM limit for your plan so you can see whether you are near it.