Turn a screen walkthrough into an agent-ready brief with narration, keyframes, written comments, and input chronology.
For Windows 10/11, macOS on Apple Silicon or Intel, and x86-64 Linux.
Download the latest release · about 1 GB because local speech recognition is included
Latest release · MIT license · CodeQL and security · Checksums · Artifact attestations · Report a vulnerability privately
FeedbackRecorder keeps the evidence and the explanation together:
- give a coding agent your narration, the relevant visuals, written notes, and a privacy-filtered chronology of clicks and keyboard activity;
- catch meaningful small changes such as a menu or dialog without producing screenshot spam;
- work by voice, by comments attached to individual keyframes, or with written instructions only;
- capture, transcribe, select frames, and store the package locally, then explicitly choose whether to paste a prompt or export a zip.
The app contacts GitHub for update metadata and for a release download only when you choose to update. You choose where to paste or export the result. The default zip excludes the raw video and your recorded voice, but it still contains sensitive context and must be inspected before sharing. See Privacy for the exact data flow.
- Download and install the build for your platform.
- Record a screen or import a video, then frame the region the agent should inspect.
- Add Notes for the agent and review screenshots if useful. Choose On this computer to copy a short prompt, or Attach to a chat to drag the prepared ZIP into your conversation.
New here? Follow Getting started for the complete first handoff.
The main window moves away and leaves a compact controller with elapsed time, microphone level, Pause/Resume, and a Stop button whose menu also offers discarding the take. Pause mutes the microphone and leaves the interruption out of the recording.
Drag over the stable region the agent should inspect, then scrub through the video to confirm the frame still fits.
Review › on the screenshot row opens screenshots to enlarge and annotate. Copy prompt points to the saved package; Copy as text copies the notes, narration, visual references, and input chronology together. The full prompt stays hidden until Show prompt is chosen. Attach to a chat offers a draggable ZIP, with optional Copy message and Save ZIP… below the card.
Capture
- Record a display, or an app window on macOS, with optional microphone narration.
- Import an existing video instead of recording.
- Use a compact controller and a global stop shortcut while recording.
- Pause and resume the same recording with the microphone muted during interruptions.
Context
- Transcribe narration locally in a selected language or with automatic detection.
- Select keyframes when the screen meaningfully changes and identify revisits without saving duplicate images.
- Frame and reframe the relevant region, enlarge keyframes, and attach per-frame comments or general written instructions.
- On macOS and Windows, optionally timestamp clicks, scrolling, shortcuts, navigation keys, and counts of ordinary typing; typed characters are not stored.
Handoff
- Copy a short prompt, a suggested message for a chat, or the whole brief as text.
- Drag or save a zip containing the brief, transcript, screenshots, notes, input
chronology, and
run.json. - Add raw video and narration audio only when explicitly selected.
- Browse and search earlier recordings, reopen one to edit its comments or copy it again, or move it to the trash.
Cross-platform, appearance, and updates
- Windows 10/11, macOS on Apple Silicon and Intel, and x86-64 Linux.
- Light, dark, or system appearance.
- Settings groups language and the clicks/keys switch under Recording, and the save folder and appearance under General. The version and Check for updates sit quietly at the bottom; newer releases get a single-action banner.
- Update checks use GitHub release metadata; installer downloads start only after confirmation and are checksum-verified before use.
Privacy
- Capture, transcription, keyframe extraction, notes, and input activity are processed and stored locally.
- No account, advertising, analytics, or telemetry.
- The default location avoids known cloud-sync roots when possible and warns when a selected folder appears synchronized.
- Exports remain sensitive: screenshots, transcripts, notes, input chronology,
and
run.jsoncan expose private content or local paths.
FeedbackRecorder is open source and published by Rabamba AB. main
is protected by pull request checks. CI includes tests on supported operating
systems, Gitleaks and secret scanning, dependency review, and separate CodeQL
analysis. Those controls reduce risk; they do not prove that the software has
no vulnerabilities.
Build-time vendor inputs use immutable revisions plus expected sizes and SHA-256
digests. Releases include
SHA256SUMS.txt
and GitHub artifact attestations:
gh attestation verify <downloaded-file> -R magnuslandahl/FeedbackRecorderSee Shipped components for bundled dependencies and report vulnerabilities privately.
macOS releases are signed with the Rabamba AB Developer ID and notarized by Apple, so they open normally. Windows installers are unsigned, so Windows shows a first-run warning and managed devices may block installation. Checksums and attestations verify bytes and GitHub build provenance; they do not replace platform signing. See Code signing and notarization.
Choose the build for your platform. The download is about 1 GB because it includes the speech-recognition model used for local transcription.
| Platform | Download | First launch |
|---|---|---|
| Windows 10 or 11, x64 | FeedbackRecorder-Windows-x64-Setup.exe | Open the installer. At Windows protected your PC, choose More info → Run anyway, then finish setup. |
| Mac, Apple Silicon (M1 or newer) | FeedbackRecorder-macOS-arm64.dmg | Drag the app to Applications and open it. Confirm Open when macOS asks about an app downloaded from the internet. |
| Mac, Intel | FeedbackRecorder-macOS-x64.dmg | Same as Apple Silicon. |
| Linux, x86-64 | FeedbackRecorder-Linux-x86_64.AppImage | Run chmod +x FeedbackRecorder-Linux-x86_64.AppImage, then ./FeedbackRecorder-Linux-x86_64.AppImage. |
On macOS, allow Screen Recording and Microphone when prompted. Optional click and keyboard chronology uses Accessibility and Input Monitoring. Use the app's restart button after changing permissions so macOS applies them to the new process.
If Windows offers no Run anyway, Smart App Control or an organization policy may be blocking unsigned apps. If macOS refuses to open the app or reports that it is damaged, follow the recovery steps in Getting started.
- Getting started — download choice, installation, permissions, and the first handoff
- User guide — complete capture, framing, annotation, export, settings, and troubleshooting workflow
- Privacy — canonical data handling and network behavior
- Signing — how releases are signed and notarized
- Troubleshooting
- Contributing and Code of Conduct
- Developer architecture
- MIT license
The repository's social preview is
docs/images/social-preview.png (1280 × 640),
regenerated with npm run shots:public in app/. GitHub does not read it from
the repository: after it changes, an administrator uploads it again under
Settings → General → Social preview.



