Holy shit Kyle. I had no idea you were working on this. This is amazing. Your patch is also very instructive on what you need me to do for you to make this more reasonable.
I'm guessing that performance of this relative to xterm right now isn't... the best, mainly because the way you're grabbing the viewport seems expensive. I'm curious though if you did any benchmarks?
One thing you probably really want to expose is the new RenderState API: https://github.com/ghostty-org/ghostty/blob/main/src/termina... You're doing per row line grabbing currently which is probably pretty slow. The RenderState API is stateful and produces the state necessary to create a high-performance, delta update renderer. It's what our production GPU renderers are now built on (but the API itself is compatible with any kind of renderer). It'd be better for you.
After all that, I'm very curious even at this rudimentary level what the performance on various benchmarks look like compared to xterm.js.
Curious to know what makes this "a proper VT100 implementation in the browser, not a JavaScript approximation of one" -- isn't Ghostty also an approximation, just implemented in a different language? Seems unnecessarily pejorative to me.
Aren't terminals also called... terminal emulators? All modern terminals would be an approximation by this logic. Some approximate backwards compatibility with VT** spec more than others.
I don't mean to derail discussion about a cool project, but it still seems to imply xterm.js is somehow "improper" emulation (though I might be misreading it).
Terminal emulators are all approximations of terminals, regardless of the programming language.
Ghostty is so great. Cross-platform but native on Mac and Linux. Core written in a cool random language, showing that you can have well-behaved Mac apps that aren’t just pure Swift / Objective C. Same great design no doubt helps here.
When users reattach to their session we render the terminal state and output from ghostty. Super cool and works really well. It’s basically tmux-lite in 1k LoC
I like that. I don’t want to use tmux (and I don’t when I’m working on my local machine), but I can’t escape it when SSHing. I could ssh to a ton of sessions, but then I 1. Have to remember their names 2. Can’t easily create a new pane (on the remote host) for some short task and 3. Need yet another solution for restoring my pane layout for when my client restarts.
Maybe I’ll try the session name thing, I just foresee it being annoying. Do you see your tool as a shpool replacement?
I also (coincidentally) just started using OP's coder, and that also sets up ssh config to use special wildcard hosts, and unfortunately 'hogs' the config (it threatens to trash any changes to the coder section).
Always love seeing new vt100 implementations! Here's my current pet one running inside Unreal5:
https://imgur.com/a/2vsGS0G
Vibe coding with Claude in a UE5 Editor tab is a lot of fun ... especially once you give it access to the Remote Control API in UE5, it can control an avatar, take screenshots. All very helpful for debugging 3d games:
I haven't really looked at MCP because it sounds like it is little bit broken?
The Remote Control API is just HTTP/JSON so Claude wrote some powershell scripts to query objects from the endpoint.
We gave Claude a Character Actor in game with an AI controller attached, and it can call functions to the AIController - MovetoXYZ(), Teleport(), TakeScreenshot() etc
This is awesome! I'm Syrus, from Wasmer. Would love to help you with this!
We are releasing soon a new version of wasmer-js, so it should be very easy to use it with webassembly.sh (note webassembly.sh and wasmer.sh share the same code)
Everything went smooth (just added a new comment on top of this thread for visibility!), only nit is that `convertEol` didn't work, so I had to manually convert `\n` to `\r\n`.
> Hint: Wasm lacks an MMU, meaning that Linux needs to be built in a NOMMU configuration. Wasm programs thus need to be built using -fPIC/-shared. Alternatively, existing Wasm programs can run together with a proxy that does syscalls towards the kernel. In such a case, each thread that wishes to independently execute syscalls should map to a thread in the proxy. The drawback of such an approach is that memory cannot be mapped and shared between processes. However, from a memory protection standpoint, this property could also be beneficial.
Ghostty is a terminal like iTerm. This compiles it so it runs in the browser directly, or browser-based environments like VS Code or the Hyper terminal. Without that you’d have to reimplement a whole terminal in JavaScript. Which is what people have been doing with via the xterm.js project. Naturally, there is effort and bugs that go into maintaining a clone/port like that. This lets you use the Ghostty terminal code directly - compiled to WebAssembly and with no other dependencies - as an API-compatible drop-in replacement
That actually pretty much is the ELI5. Its merely a different terminal that offers more features than iTerm2 and also runs on OSX.
Unless you actually need/want those features (which, although I am a terminal aficionado, I must admit are niche as fuck), pick whichever terminal makes you happy. Features that are important to some people are performance, Unicode support, and OS support.
You could argue whether or not it's a "feature", but one of the thing ghostty claims as an advantage is the out of the box configuration.
With no config at all, ghostty looks nicer than my alacritty setup. The rendering is just real nice. I could probably get alacritty to look as nice or nicer, but ghostty just worked this way with no config needed.
So you could consider aesthetics and rendering quality, and simplicity of setup both as features, which people may need/want (or not).
I wouldn't argue against that at all: OOBE is absolutely a feature.
Problem is, we don't all agree with what the OOBE should be. I, for example, always strip out menus, tabs, and other UI features. For me, the terminal that requires the least lines in the config file is probably going to be the winner (assuming no unfixable defects that effect me).
I also have ton of questions. Hopefully the author can add more documentation to ghostty. Right now I don't fully understand the use cases or how people may benefit from ghostty.
This runs in the browser, so it would allow you to connect to a server from your browser and render normal terminal commands in that environment
For instance if you're a cloud provider, and you want people to be able to "drop in a shell" on a machine, but make that available through the web, you could use this
This is super cool! I'm close to releasing a project to make Ink work in the browser for building cross-platform terminal apps. (Ink is what Claude Code / Gemeni CLI use for rendering.)
Currently it uses Xterm.js - but I'll have to try swapping Xterm out for ghostty-web!
I got ghostty-web working with this and it is great. Unfortunately it seems I don't have enough karma to post a link to it. But it's ink-web on npm if you're interested!
The nice thing about this is that you do not need an OS / backend of any kind - everything runs in the browser - similar to webcontainers but without even needing a container.
I always thought it would be interesting in backend systems to catch a certain exceptions and auto-generate a link to a shell. Given the proper authentication is implemented would this be a good tool to achieve that "remote debug" shell?
Search had been a blocker, but that's coming soon; beyond that not sure that there's any reason other than inertia. Alacritty is totally fine, but excited for the Ghostty focus on performance (obviously), and the font rendering stuff looks excellent (though TBD how much of that we can "just use" vs needing to copy-pasta)
i switched from alacritty -> ghostty, and occasionally use zed.
i can't recall why i made the transition (maybe just to try it out, and it became default?). i can't think of any practical consequences of this transition.
why do you care about whether zed uses alacritty or ghostty under the hood?
Kitty Graphics Protocol support and subtle font rendering differences between Ghostty and Alacritty that drive me nuts.
I have reported font rendering issues to Alacritty in the past and let's just say the developer was not exactly receptive to fixing them since they occur on macOS and not his preferred OS of Linux.
You could use Firefox and then set the accelerator key to Super, changing all application bindings to use Super instead of Control. This is my setup and it’s great.
Does this version support custom GPU shaders? I have been looking for a command-line with cool-retro-term aesthetics that can run in the browser for a while.
It's maybe possible. Custom shaders are OpenGL syntax so it'd require transforming them to something compatible with WebGL/WebGPU. In Ghostty GUI we use spire-cross and glslang to transpile shaders at runtime from OpenGL to Metal or OpenGL (with features our host supports).
We'd have to look and see if these support WebGL/GPU. The next problem would be making all that fit into the wasm blob.
Or, we may be able to skip most of this is the OpenGL syntax we use is already compatible. Then no transpiling necessary...
Oh it's available, but it only works for about 5 minutes and then it becomes so slow that it's unusable. There was a very long issue about it that was closed with seemingly no effort on behalf of the maintainers.
omg! Seriously, wow. That was quick! Only just heard that Hashimoto libbed out his terminal the other day and remarked about how smart that was... And it made this possible
We can easily make a browser shell that let's people run basic commands, but presumably most want to try `vim` and other commands they'd typically invoke.
If you have a Dockerfile that bundles ghostty-web with a backend (even just `ttyd` or a simple socket server), I’d love to host the official demo for you. We can give it a dedicated isolated environment so people can run vim safely.
`npx @ghostty-web/demo@next` starts an HTTP server on `localhost:8080`, so you could just wrap a basic Dockerfile with NPM installed (and maybe a variety of fun Linux tooling, ala vim).
Feel free to shoot me an email: [email protected]. I'll happily add it to the README.
The comparison has two points one of which says that for "Complex scripts (Devanagari, Arabic)", ghostty-web has " Proper grapheme handling". But when I tried something extremely simple at the https://ghostty-web.wasmer.app/ or https://ghostty.ondis.co/ demos (just `echo "कवि"`), it failed pretty miserably (and in different ways in the two demos). So I'm not sure what to make of it.
(Seems slightly better in the actual Ghostty app, but not in any usable way: here was my comparison last year https://shreevatsa.net/post/terminal-indic/ and Ghostty is comparable to the other terminal apps, better than some, worse than some. Anyway the fact that it works in Ghostty but not in ghostty-web casts doubt on “Ghostty's emulator is the same battle-tested code that runs the native Ghostty app”. Do you understand what is going on?)
I'm guessing that performance of this relative to xterm right now isn't... the best, mainly because the way you're grabbing the viewport seems expensive. I'm curious though if you did any benchmarks?
One thing you probably really want to expose is the new RenderState API: https://github.com/ghostty-org/ghostty/blob/main/src/termina... You're doing per row line grabbing currently which is probably pretty slow. The RenderState API is stateful and produces the state necessary to create a high-performance, delta update renderer. It's what our production GPU renderers are now built on (but the API itself is compatible with any kind of renderer). It'd be better for you.
After all that, I'm very curious even at this rudimentary level what the performance on various benchmarks look like compared to xterm.js.
Excellent work!
I'll swap it over to the new RenderState API and post some benchmarks!
Many kudos to y'all, we were shocked how simple it was to hack this together.
Terminal emulators are all approximations of terminals, regardless of the programming language.
I'm probably also just taking things personally.
It will be used to render ANSI inside opencode
https://github.com/remorses/ghostty-opentui
https://github.com/emadda/hot-notes/
https://ghostty-web.wasmer.app/
How to try it locally:
Source code: https://github.com/wasmerio/webassembly.sh (updated from xterm into ghostty-web without any issue!)When users reattach to their session we render the terminal state and output from ghostty. Super cool and works really well. It’s basically tmux-lite in 1k LoC
Maybe I’ll try the session name thing, I just foresee it being annoying. Do you see your tool as a shpool replacement?
The Remote Control API is just HTTP/JSON so Claude wrote some powershell scripts to query objects from the endpoint.
We gave Claude a Character Actor in game with an AI controller attached, and it can call functions to the AIController - MovetoXYZ(), Teleport(), TakeScreenshot() etc
Currently you need the command-line to try it, which is an unfortunate UX.
We are releasing soon a new version of wasmer-js, so it should be very easy to use it with webassembly.sh (note webassembly.sh and wasmer.sh share the same code)
https://github.com/wasmerio/wasmer-js/tree/main/examples/was...
Have you seen container2wasm or ktock/vscode-container-wasm?
container2wasm: https://github.com/container2wasm/container2wasm
ktock/vscode-container-wasm: https://github.com/ktock/vscode-container-wasm
ktock/vscode-container-wasm-gcc-example: https://github.com/ktock/vscode-container-wasm-gcc-example
From joelseverin/linux-wasm: https://github.com/joelseverin/linux-wasm :
> Hint: Wasm lacks an MMU, meaning that Linux needs to be built in a NOMMU configuration. Wasm programs thus need to be built using -fPIC/-shared. Alternatively, existing Wasm programs can run together with a proxy that does syscalls towards the kernel. In such a case, each thread that wishes to independently execute syscalls should map to a thread in the proxy. The drawback of such an approach is that memory cannot be mapped and shared between processes. However, from a memory protection standpoint, this property could also be beneficial.
https://ghostty.ondis.co/
Stellar work, Kyle! Cheers
https://github.com/coder/ghostty-web/pull/76
Unless you actually need/want those features (which, although I am a terminal aficionado, I must admit are niche as fuck), pick whichever terminal makes you happy. Features that are important to some people are performance, Unicode support, and OS support.
The most decidedly non-ELI5 feature comparison: https://www.jeffquast.com/post/state-of-terminal-emulation-2... and https://ucs-detect.readthedocs.io/results.html
With no config at all, ghostty looks nicer than my alacritty setup. The rendering is just real nice. I could probably get alacritty to look as nice or nicer, but ghostty just worked this way with no config needed.
So you could consider aesthetics and rendering quality, and simplicity of setup both as features, which people may need/want (or not).
Problem is, we don't all agree with what the OOBE should be. I, for example, always strip out menus, tabs, and other UI features. For me, the terminal that requires the least lines in the config file is probably going to be the winner (assuming no unfixable defects that effect me).
For instance if you're a cloud provider, and you want people to be able to "drop in a shell" on a machine, but make that available through the web, you could use this
Currently it uses Xterm.js - but I'll have to try swapping Xterm out for ghostty-web!
The nice thing about this is that you do not need an OS / backend of any kind - everything runs in the browser - similar to webcontainers but without even needing a container.
I wonder if https://github.com/zed-industries/zed/discussions/18129 is still accurate. Would love to be able to use Ghostty as my Zed terminal.
Search had been a blocker, but that's coming soon; beyond that not sure that there's any reason other than inertia. Alacritty is totally fine, but excited for the Ghostty focus on performance (obviously), and the font rendering stuff looks excellent (though TBD how much of that we can "just use" vs needing to copy-pasta)
(Edit) Download it here: https://github.com/ghostty-org/ghostty/releases/tag/tip
i can't recall why i made the transition (maybe just to try it out, and it became default?). i can't think of any practical consequences of this transition.
why do you care about whether zed uses alacritty or ghostty under the hood?
I have reported font rendering issues to Alacritty in the past and let's just say the developer was not exactly receptive to fixing them since they occur on macOS and not his preferred OS of Linux.
You can also play around with a real shared containerized machine here: https://ghostty.ondis.co/
Considering the native Ghostty does, I _think_ the answer would be yes? I might tinker around with this and let you know.
We'd have to look and see if these support WebGL/GPU. The next problem would be making all that fit into the wasm blob.
Or, we may be able to skip most of this is the OpenGL syntax we use is already compatible. Then no transpiling necessary...
[1] https://github.com/mausimus/ShaderGlass
[2] https://github.com/Swordfish90/cool-retro-term
From "WebAssembly (WASM) arch support for the Linux kernel" (2025) https://news.ycombinator.com/item?id=45784329 :
> JupyterLite still lacks a Terminal e.g. with BusyBox Ash in WASM, with a file system integrated with the Jupyter-xeus kernel file system.
> This [joelseverin/linux-wasm] appears to load much more quickly than other Linux and I think even just bash in WASM demos I've seen.
We can easily make a browser shell that let's people run basic commands, but presumably most want to try `vim` and other commands they'd typically invoke.
`npx @ghostty-web/demo@next` starts an HTTP server on `localhost:8080`, so you could just wrap a basic Dockerfile with NPM installed (and maybe a variety of fun Linux tooling, ala vim).
Feel free to shoot me an email: [email protected]. I'll happily add it to the README.
https://github.com/olson-dan/rustzork
Ghostty has much better VT100 compatibility. It should have much better performance as well once we optimize.
(Seems slightly better in the actual Ghostty app, but not in any usable way: here was my comparison last year https://shreevatsa.net/post/terminal-indic/ and Ghostty is comparable to the other terminal apps, better than some, worse than some. Anyway the fact that it works in Ghostty but not in ghostty-web casts doubt on “Ghostty's emulator is the same battle-tested code that runs the native Ghostty app”. Do you understand what is going on?)
Thank you for letting me know!