---
id: feedback
title: Report a bug without doing homework
area: Going deeper
order: 92
summary: How /feedback investigates a problem, writes the report for you, keeps your private notes out of it, and tells you when the fix ships.
---

Most bug reports never get written. Something misbehaves, you are mid-task, and writing it up properly means gathering versions and error messages you do not have to hand. So the problem goes unreported and stays unfixed for you and everyone else.

`/feedback` (new in v1.82.0) removes the homework. Say "report this" when something goes wrong, and Dex investigates on your machine, writes the report itself, shows it to you, and sends it with a reference number. When the fix ships, your next session opens with the good news and which version has it.

## What a report contains, and what it never contains

Reports are built from a fixed list of safe ingredients: Dex's version, which feature misbehaved, and error details from Dex's own workings. That list is the whole list.

Your notes, meetings, people, tasks, and conversations are never part of a report. There is no setting that changes this. Dex also shows you the exact report before anything is sent, and keeps a copy of every send attempt on your machine (in `System/.dex/feedback-log.jsonl`), so you can always see precisely what left your computer and when.

## Reporting something

When Dex misbehaves, say so in the moment:

```prompt Report this
That went wrong. Report it with /feedback — show me the report before it goes.
```

Dex investigates, drafts the report, and shows it to you. Approve it and you get a reference like **DEX-142**. The Doctor checkup (`/dex-doctor`) offers the same thing when it finds something that is genuinely a Dex problem rather than a setup problem on your machine.

By default you review every report before it goes. Once you have seen a few and trust what they contain, you can tell Dex to send future ones automatically; you can switch back any time by asking.

## The first report asks you to connect, once

Sending feedback requires signing in, by design, so that fixes can find their way back to the person who reported the problem. The first report walks you through it: a page opens at heydex.ai, you sign in, create a short code, and paste it into the terminal. It takes about thirty seconds and never needs doing again. It is the same connection the DexDiff publishing flow uses, so if you already publish diffs you are already connected.

## Hearing back

You do not need to chase anything. When a release fixes your bug, your next session opens with a note: which report, which version has the fix, and any comment from the team. Run `/dex-update` when you are ready and you have it.

If the team needs one more detail to reproduce the problem, Dex relays the question to you, gathers the answer with your approval, and sends it back the same safe way.

To check in the meantime:

```prompt What happened to my report?
What's the status of my feedback reports?
```

That shows every report this account has filed, with its current state: received, being looked at, fixed in a version, or waiting on an answer from you.

## If something looks off

Every send attempt, successful or not, is recorded in `System/.dex/feedback-log.jsonl`, and each report keeps its own file under `System/.dex/feedback/`. If a send fails, Dex says so plainly rather than pretending it went; try again later or ask Dex to show you the log.