BongoCat 1.1.0 turns global input into a Live2D pet
BongoCat v1.1.0 keeps a transparent character above the desktop and maps keyboard, mouse, or gamepad activity to its motions. Users can import their own model, and the README links a converter for models made for Bongo-Cat-Mver. The point is visual feedback and decoration, not pet care or conversation. It can work offline, and the project says it does not collect user data. Update checks are optional in the app's settings.
The 5.4 MB checkout combines a Vue interface with a Rust and Tauri desktop shell. PixiJS and a Live2D package handle the model side, while platform code listens for input and manages the transparent, always-on-top window. That architecture gives BongoCat more reach than its Windows-only inspiration. It also means source contributors cross several toolchains, and bugs can come from the webview, window manager, global-input library, graphics stack, or application state.
Version 1.1.0 covers three OS families, but Linux means X11
Version 1.1.0 has release files for Windows, macOS, and Linux across several processor types. The download guide requires Windows 10 or newer and macOS 12 or newer. Linux packages include deb, RPM, AppImage, and an AUR route, but the support line says X11. Wayland users are outside the documented target even if a package manager lets them install the application. That distinction matters on current Linux desktops where either display system may be selected.
The Linux caveat is backed by live reports. Issue 943 describes v1.1.0 aborting on Ubuntu 22.04 under X11, with a later commenter reproducing the failure using the official deb and reporting exit code 134. Issue 1028 reports a white window on Arch Linux with KDE and X11. These reports do not establish that every Linux machine fails. They do make a pre-deployment trial necessary before relying on the pet for a stream or daily workstation setup.
What happened when we ran it
Our sandbox install failed after 27 seconds with exit code 1. We ran commit 44f44bc in an unprivileged lab-node:22 container with 3 CPUs, 8 GB of RAM, and no secrets. The checkout had 294 files and about 4,774 source lines. Dependency installation did not reach a successful state, so the supplied measurement has no package count, disk total after install, build timing, or test outcome to report.
The last log lines were precise. pnpm ignored build scripts for @parcel/watcher@2.5.6, esbuild@0.25.12, esbuild@0.27.7, and simple-git-hooks@2.13.1, then raised ERR_PNPM_IGNORED_BUILDS. Its help text said to run pnpm approve-builds and choose which dependencies may execute scripts. The lockfile passed the sandbox's supply-chain policy, but that did not override pnpm's script approval. We stopped at the recorded failure and cannot say whether approval would produce a working desktop build. The checkout had 3 CI workflow files, no Dockerfile, and no tests directory.
Input monitoring can collide with the keys it watches
BongoCat v1.1.0 needs to observe system input before its model can react. Its English UI tells macOS users to add the app under Privacy and Security, Input Monitoring, then restart. Windows guidance recommends administrator mode for some system-level keys and events. Those requests fit the feature, but they give the pet unusually broad visibility for decorative software. Users should obtain binaries from the project release page and decide whether the requested access matches their comfort level.
Open issue 1032 reports that BongoCat interferes with Ctrl-plus-number grouping in Age of Empires IV on Windows, with normal behavior returning after the app closes. Issue 1010 says v1.1.0 only displays the most recently pressed key in a multi-key sequence, then fails to restore an earlier held key. The second report is a visualization error, while the first describes interference with another application. Game-specific shortcut checks belong in acceptance testing, especially when administrator mode is involved.
An open 20-minute memory report tempers the lightweight image
Issue 39 began with a report that the main process rose from 132 MB to more than 500 MB after about 20 minutes. Later comments span Windows and macOS observations, upstream input-listener discussion, and attempts to reduce the cost. The maintainer asked users to retest v1.1.0 and said it should greatly ease high usage, but the issue stayed open and received another update on September 5, 2026. That history supports profiling the current binary instead of repeating either the original figure or the hoped-for fix as universal.
Chinese documentation sits beside an English app locale
The current tree includes an en-US interface locale, yet the main README, download guide, contribution guide, and v1.1.0 release notes are Chinese. An open pull request proposes English documentation, which means an English README is not part of the current default branch. End users can use the translated settings. English-only contributors will have to translate the build and contribution material or read the source and Tauri documentation directly.
23,046 stars come with current pushes and 96 open items
GitHub showed 23,046 stars, 96 combined issues and pull requests, and a last push on September 8, 2026. New and recently updated items covered touchpad gestures, a Rust rewrite proposal, keyboard behavior, Linux windows, and macOS workspace handling. The queue is active rather than abandoned, and the combined count must not be mistaken for 96 confirmed defects. It does show that platform integration continues to generate work after the v1.1.0 release.
The latest release was v1.1.0 on April 20, 2026, while repository work continued into September. That gap alone is not a neglect signal because pushes and issue discussion are current. For an ordinary user, the release packages are easier to judge than the source tree: install one on a supported system, verify the permissions, try every important shortcut, and watch resource use through a long session. Source contributors face the additional pnpm approval failure that ended our run.