> ## Documentation Index > Fetch the complete documentation index at: https://docs.coderabbit.ai/llms.txt > Use this file to discover all available pages before exploring further. # Work with findings in JetBrains IDEs > Browse CodeRabbit findings in the JetBrains tool window and editor, compare and apply suggestions safely, ignore or reopen findings, send fixes to Claude Code, Codex CLI, or opencode with Fix with AI, give feedback, and revisit saved reviews. Open Beta (The CodeRabbit plugin for JetBrains IDEs is in beta.) When a review runs, its findings appear in the **CodeRabbit** tool window, and open findings also appear as markers in the editor. From a finding, you can go to the code, apply the suggested change, hand the fix to a coding agent, or ignore it. ## Browse findings The results area has a header and two panes: * The header shows the **Review** menu for [saved reviews](#review-history), how many findings are resolved, and the **Fix all**, **Review again**, and **New review** buttons. * The list groups findings by severity: **Critical**, **Major**, **Nitpick**, and **Info**. Minor and trivial findings both appear under **Nitpick**. Each finding shows its title and its file and line. * The **Finding details** pane shows the finding you select: its severity, title, location and status, CodeRabbit's comment, the **Affected code**, and the **Suggested fix** or **Fix guidance**. To narrow the list, select **Filters**, then filter by status (**Open** or **Resolved**), by severity, or by text in titles, details, and file paths. **Resolved** covers findings you applied or ignored. **Clear filters** shows every finding again. The details pane has **Apply suggestion**, **Fix with AI** (or **Copy fix prompt**), and **More actions**, which holds **Open in editor**, **Show diff**, **Ignore**, **Reopen**, and **Feedback**. ## Go to a finding in the editor To open the code, select the location link in **Finding details**, or choose **More actions → Open in editor**. Selecting a finding in the list only shows its details. Every open finding of the review shown in the tool window also gets a gutter icon, a highlight on its lines, and a mark on the editor's scroll bar. Hover over the gutter icon to see the finding's title, and select it to open a popup with the details and the **Show diff**, **Apply suggestion**, **Fix with AI**, and **Ignore** buttons. Markers go away when you apply or ignore the finding. The editor uses the IDE's error, warning, and information styles: **Critical** findings show as errors, **Major** findings and minor nitpicks as warnings, and trivial nitpicks and **Info** findings as information. **Show diff** opens a diff window that compares the **Reviewed code** with the **Suggested code**. It shows the code as it was when CodeRabbit reviewed it, not your current file. ## Apply a suggestion **Apply suggestion** replaces the reviewed lines with CodeRabbit's suggested code as a single edit, and marks the finding **Applied**. It doesn't save the file, so you can check the change first. Undo (`Ctrl+Z` on Windows and Linux, `Cmd+Z` on macOS) reverts the edit and reopens the finding. The plugin applies a suggestion only when it can do so safely. Otherwise it shows why: | Message | Cause | What to do | | - | - | - | | "The file changed in this range after the review. Re-run the review before applying." | The lines are no longer the ones CodeRabbit reviewed. | Select **Review again**, or use **Fix with AI**. | | "This file is read-only. Make it writable before applying the suggestion." | The file is read-only. | Make the file writable, then try again. | | "This review includes guidance, but no directly applicable code change. Use the fix action to continue." | The finding has advice but no replacement code. | Use **Fix with AI**. | | "Apply is unavailable because this saved review lacks the reviewed code snapshot. Run a new review to apply a suggestion safely." | The review doesn't hold the reviewed lines, for example because the file was larger than 2 MB. | Run a new review. | If a finding has more than one suggestion, the plugin applies the first one. You can also apply suggestions from a [saved review](#review-history), as long as the lines haven't changed since. ## Ignore or reopen a finding **Ignore** marks a finding **Ignored**, and **Reopen** marks an ignored or applied finding **Open** again. There's no separate resolve action: a finding counts as resolved once you apply or ignore it. These statuses stay on your computer, saved with the review. CodeRabbit doesn't receive them, they don't change your pull request reviews, and a later review can report the same issue again. To tell CodeRabbit that a finding is wrong, [give feedback](#give-feedback-on-a-finding). ## Fix with AI **Fix with AI** hands a finding to a coding agent. The **Fix findings with** setting in **Settings → Tools → CodeRabbit** picks the agent: | Option | What **Fix with AI** does | | - | - | | **Copy to clipboard** | Copies CodeRabbit's fix prompt, so you can paste it into any coding agent or AI chat. This is the default, and the button reads **Copy fix prompt**. | | **Claude Code** | Opens a terminal tab named "CodeRabbit - Claude Code" in the project folder and runs `claude` with the prompt. | | **Codex CLI** | Opens a terminal tab named "CodeRabbit - Codex" and runs `codex` with the prompt. | | **opencode** | Opens a terminal tab named "CodeRabbit - opencode" and runs `opencode run` with the prompt. | For the three agents, the plugin builds a prompt from the finding's title, comment, guidance, and suggested code, and tells the agent to check the finding against the current code before it edits. The prompt is copied to the clipboard too, so you can paste it if the agent doesn't start. Each **Fix with AI** opens a new terminal tab. Claude Code and Codex start an interactive session with the prompt, so you can follow up in the same tab; opencode runs the prompt once and exits. The plugin adds no permission flags, so the agent's own approval prompts and settings apply. The agent must be installed and on the `PATH` of the shell that the IDE's terminal uses; in Remote Development, that's the host. The plugin doesn't check this; if the agent is missing, the terminal shows the shell's "command not found" error. On macOS and Linux, a long prompt goes through a temporary file named `coderabbit-fix-*.txt`, which is deleted when the agent exits without an error. Fix with AI doesn't change a finding's status. After the agent finishes, select **Review again** to check the fix. Fix with AI sends nothing to CodeRabbit; your agent's own terms apply to what you send it. It's also separate from CodeRabbit's [Coding Agent](https://docs.coderabbit.ai/code/index.md), which runs tasks in the cloud. ### Fix several findings at once **Fix all** opens "Select findings to fix", which lists the open findings with all of them selected. Clear the ones to skip, then select the button, such as **Send 4 findings to Claude Code** or **Copy prompt for 4 findings**. You can send up to 50 findings at a time. ## Give feedback on a finding To tell CodeRabbit that a finding is wrong, choose **More actions → Feedback**. In "Send finding feedback", pick a **Reason** (**Not helpful**, **Incorrect**, **Poorly formatted**, **Wrong lines**, or **Suggestion is invalid**), optionally add details of up to 4,000 characters, and select **Send feedback**. The feedback includes the finding's file path and comment. ## Review history The plugin saves each completed review of a repository on your computer, and keeps the newest 20. Cancelled, failed, and skipped reviews aren't saved. To open a saved review, pick it from the **Review** menu in the results header. Saved reviews are listed by date and branch, and **Current review** returns to the latest one. After you reopen a project, the tool window shows the newest saved review once you're signed in. In a saved review, you can still apply, ignore, reopen, fix, and give feedback on findings. Saved reviews hold the findings, their statuses, and the reviewed lines, but not full files. They're stored in the project's workspace file, usually `.idea/workspace.xml`, and nowhere else. If that file is committed to version control or included in a shared backup, others can read the saved reviews, so keep it out of both. In Remote Development, they're stored on the remote host. To delete them, choose **Clear local review history…** from the **Options** menu (⋮) in the tool window's title bar, then **Clear history**. This deletes the saved reviews of the current repository only, and can't be undone. Signing out, switching organizations, and uninstalling the plugin don't delete them. ## What's next > [card] [Run a review](https://docs.coderabbit.ai/ide/jetbrains/review.md) > > Pick the changes and base branch to review, or review each commit automatically. > [card] [Privacy and network](https://docs.coderabbit.ai/ide/jetbrains/privacy.md) > > See what a review sends and what the plugin stores on your computer. > [card] [Troubleshooting](https://docs.coderabbit.ai/ide/jetbrains/troubleshooting.md) > > Fix findings that don't open, suggestions that won't apply, and agents that don't start. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.