Heya - I'm the author. Glass is a cross-platform MCP server that gives an agent a way to see and interact with GUI apps. I don't
like building UI. Agents can code it, but can't verify or debug it.
Glass exists to close that loop.
Glass is purpose-built for software development, and aims to enable headless, parallelized agent work while
the user continues to use the machine unhindered[0]. It minimizes token usage, increases consistency, and just goes faster
by preferring to interact with an app's accessibility tree, before taking the screenshot route (which glass also fully supports).
Glass is cross-platform and the code is Apache-2.0 licensed. We release builds for Linux, Windows, and macOS hosts.
Glass supports development targeting the three above host platforms, Android (all hosts) and iOS[1] (macOS hosts). The
internal API is backend-agnostic, and new platforms can be added in the future pretty easily.
Feedback is very welcome, either here or at github. If it breaks on your app, I am especially interested in
knowing about that. I hope you build great things.
[0] Due to the license enforcement on Windows and macOS, isolated headless development targeting those
platforms requires a VM. It will otherwise interact directly with the desktop.
Android and iOS development on those platforms can still be done headlessly via the emulators.
[1] The iOS backend does not currently support multi-touch gestures.
Having good error messages is one of the important priorities for me in my compiler, so I made the commitment early and am using codespan_reporting[0] to report the errors, which is going well so far.
Oh, it looks like the release command would fit the `build` step I mention.
As far as the after-deployment tasks go, we automate those just like migrations - they're (occasionally very slow) one-offs that we don't want to hold up a restart for. Really, an analogue to the release command that can run after restart is all that I'm talking about here.
My big want here is containerized deployments with build + release steps to allow me to e.g. run migrations and after-deploy tasks (we use both). This prevented a move to Render for us previously.
Only if said "reasonable person" were around in the 90's, and knew the shit that MS was trying to pull. I happen to be one of those people, but that's not the catch of the day here.
To me, it feels like the same effect as Ruby ~15-10 years ago. Lots of chaos as good (subsequently winnowed down to preferred), patterns were discovered, and then things stabilized. It isn't sexy anymore, but to this day I've yet to discover a better tool, that lets me blast out features and make changes at rapid pace, than Ruby (and Rails on the web side).
Glass is purpose-built for software development, and aims to enable headless, parallelized agent work while the user continues to use the machine unhindered[0]. It minimizes token usage, increases consistency, and just goes faster by preferring to interact with an app's accessibility tree, before taking the screenshot route (which glass also fully supports).
Glass is cross-platform and the code is Apache-2.0 licensed. We release builds for Linux, Windows, and macOS hosts. Glass supports development targeting the three above host platforms, Android (all hosts) and iOS[1] (macOS hosts). The internal API is backend-agnostic, and new platforms can be added in the future pretty easily.
Feedback is very welcome, either here or at github. If it breaks on your app, I am especially interested in knowing about that. I hope you build great things.
[0] Due to the license enforcement on Windows and macOS, isolated headless development targeting those platforms requires a VM. It will otherwise interact directly with the desktop. Android and iOS development on those platforms can still be done headlessly via the emulators.
[1] The iOS backend does not currently support multi-touch gestures.