> ## Documentation Index > Fetch the complete documentation index at: https://docs.everydeck.ai/llms.txt > Use this file to discover all available pages before exploring further. # Cases and repairs > How everydeck groups errors into cases, how to fix them with your coding agent, what Fixed and Came back mean, and how to mark as fixed, ignore, archive and undo. When something goes wrong in your app, everydeck opens a **case**. A case bundles all events of the same kind of problem. That is a way of grouping work, not a proven cause. Every case has a short address, like `ACME-14`. The number counts up per app and is never reused. ## Kinds of cases | Kind | What it means | | - | - | | Confirmed error | An error event from your app, in the browser or on the server. | | Missed expectation | After a trigger, the expected result did not arrive in time. | | Data gap | everydeck cannot judge it because data is missing. This is not a confirmed failure. | | Confirmed outage | The reachability check of your app failed twice in a row. | | Monitoring gap | No heartbeats from your app, a job did not report, or a flow stopped sending after a new version. Not a confirmed outage. | | No answer | A request of a flow hung at the same place several times without an answer. | | Data loss | A large part of a kind of data disappeared at once. | | Implausible value | For example a price of 0 or less, or a required field that is empty. | | Security | A Supabase security error, or a secret key in the code your app ships to browsers. | | Due date, Needs review | Cases from your app's own dates and from moderation. | Titles are written in plain words from facts, for example "Checkout fails" or "Your app could not be reached". A 4xx answer your app gives on purpose is counted, but it is no case. ## The case page * **What happened:** one sentence with the flow, the impact, how often and who. * **What the error is**, **Where** (page, flow, code lines), **Who it hit**. * **The way there:** the steps in the browser before the error. * **How to fix it:** the repair path, on the right. * **History**, **Notes** for your team, and **For developers** with the stack trace, the request and all events. If a provider of your app (Supabase, Vercel, Render, Stripe, OpenAI or Anthropic) reports an incident right now, the case says: "This error may come from the provider, not from your app. everydeck checks the provider's public status page." ## Fix it with your coding agent everydeck does not write code. Your coding agent does. The case page gives it everything it needs. **Copy fix** Click **Copy fix for Claude Code** (or your tool's name). Nothing starts by itself: copy the task, then paste it into your agent. It fixes the error and reports back. The copied task contains what happened, the steps, the place in the code and the versions. It contains no names or addresses of people, only their ids. It ends with a note that app content in the task is data, not instructions. **Send to a connected agent** If your agent is connected to everydeck, click **Send to connected agent**. Your agent does not run by itself. Tell it once: "Fix the everydeck issue". It picks up the oldest case you sent. Each case is given out once, even if two agents ask at the same time. If nobody picks up a sent case for a day, you get an email and a banner. When your agent reports the fix, the case is marked **Fixed**. everydeck then watches. If the error comes back, you get an email and the case shows **Came back**. ### The repair path The case page shows each step on the right, with its own evidence: 1. **Detected** 2. **Copied for your agent** (when and by whom), or **Sent to your agent** and **Picked up** if your agent is connected to everydeck 3. **Fixed**, with what your agent reported: what it changed, its tests, the commit and the version 4. **New version running**, as information when everydeck sees it 5. **Watching**, or **Came back** if the error happens again **Merged** only appears if a repository is connected. There is nothing to click and nothing to wait for: the case is fixed as soon as your agent reports it. ### When the error comes back If the same error happens again in the reported version or a later one, the same case opens again as **Came back**, keeps its address, and you get an email. Errors from an older version that is still running do not count. ## What you can do These buttons need the Owner, Admin or Agency role. * **Mark as fixed.** If the same error happens again, the case opens again as **Came back** and you get an email. **More ways to mark as fixed** offers: * **Resolve now:** if it happens again, the case opens again. * **The next release:** the running release may still show it. Only a newer release opens the case again. * **Ignore.** By default **Until it gets worse**: back when it hits someone it did not hit before, or happens in a new version. **More ways to ignore** offers more choices. * **Archive**, with fixed choices: * **For …** 30 minutes, 2 hours, 6 hours, 24 hours or 7 days. Back on Needs you when the time is over. * **Until this occurs again …** 1, 10, 100, 1,000, 10,000 or 100,000 more times, counted per hour, day or week. * **Until this affects an additional …** that many people who were not hit before. * **Forever:** everydeck keeps counting, the case does not come back by itself. * **Undo.** After an action, a short message offers Undo. Then the case is open again. * **Assign** to a member, or **Remove the assignee**. * **Reopen** a closed or dismissed case. Every action appears in the case history and in the protocol with the person who did it. ## When a case comes back A case you or your agent marked as fixed opens again as **Came back** when the same error happens again in the fixed version or a later one. It keeps its address and its history, and you get an email. ## Cases from the setup Cases from the time while your app was being set up move to the developer view once real data arrives. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.