Getting help when something breaks

Run one command and Ghosty diagnoses what went wrong — with fix steps you can hand to your coding assistant.

Getting help when something breaks

Sometimes a deploy fails, or an app that worked yesterday starts misbehaving. You don't have to figure out why on your own: Ghosty can look at what just happened on the platform side and tell you what it found.

Run ghosty report

From your app's folder, run:

ghosty report

Ghosty gathers what just happened — including the last failure the CLI saw on this app, if it was recent — asks the platform to diagnose it, and prints the verdict right in your terminal. Nothing to configure, and running it never changes your app.

You can run it any time something feels off, but it's most useful right after a failure. When a command like ghosty deploy fails, the CLI will suggest it:

✘ Deploy failed: the build did not complete.
Run `ghosty report` to diagnose.

Working outside the app folder? Point the report at an app with ghosty report --app <app-name>.

What the verdicts mean

Every diagnosis ends in one verdict:

Verdict What it means What to do
A problem in your app's code The failure comes from the app itself — for example a missing dependency or an error thrown at startup. Follow the fix steps below the verdict.
A configuration problem Something in the app's setup needs attention — an environment variable, a connected service that isn't linked, a domain that isn't ready. Follow the fix steps; most take a minute in the console or CLI.
A problem on the platform side Ghosty itself hit a problem. These reports are escalated to the Ghosty team automatically — you don't need to file anything. Usually just retry later; the report tells you if there's a workaround.
A problem with an external service Something your app depends on outside Ghosty (a third-party API, a package registry) isn't behaving. Check that service's status; retry once it recovers.
No clear cause found The evidence didn't point to a single cause. The report still lists what was checked, which narrows things down.

A diagnosis always counts as a success — even "no clear cause found" tells you what it ruled out.

Hand the fix to your coding assistant

When the verdict points at your app's code or configuration, the report prints numbered fix steps written so a coding assistant can act on them directly. Copy the whole report output into your assistant (Claude Code, Cursor, or any other) and ask it to follow the steps — they name the exact commands and files involved.

If your assistant drives the CLI itself, it can run the report in machine-readable form:

ghosty report --json

This prints the full diagnosis as JSON, ready for an agent to consume.

Even better: connect your assistant to Ghosty directly and it can diagnose and fix without any copy-pasting — see Connect your coding assistant.

Good to know

  • Reports are per app. Run the command in the app folder you're asking about (or use --app).
  • Recent failures attach automatically. A failure from the last 30 minutes is included as context, which makes the diagnosis sharper. It's only attached when it belongs to the same app and workspace you're asking about.
  • There's a daily limit on diagnoses per app. If you hit it, the report says so — try again the next day.
  • If your workspace doesn't have diagnosis yet, the report tells you an update is coming. In the meantime, ghosty deploys and ghosty logs show what happened.