I strongly resonate with the problem statement, but this implementation was very far off the mark for me. Every interaction feels bad.
I fired it up, and the first thing I notice is that the arrow keys don't work. I can't select Claude Code. Oh, apparently it's in a different control, so I have to press Tab, and then the arrow keys work. Wow, this list of buttons has a slow scrolling animation when navigating it. Can I turn that off? Press enter on Claude, now I'm in a tiny modal window. Press enter, because I want to do the obvious thing, but apparently the obvious thing is "show in launcher", so the background of the modal is weirdly changing while a tiny single character inside the button is indicating that this is the part I'm supposed to be focusing on. No, I want to do the obvious thing of running Claude code. You could easily fit the 4 actions of this form on my screen, but by choosing to use a tiny modal window you're now forcing yourself to use another modal drop-down control to choose the action and a separate "yes actually do it" button, so the OBVIOUS ACTION of RUN THIS AGENT requires pressing tab, enter, down, down, down, enter, tab, enter. Great. Now I'm at a chat interface with an error screen, because it isn't installed. Quit the program, restart, enter, tab, enter, down, down, enter, tab, enter to install. It shows a successful run of the "ACP adapter" for claude. Shift-tab, enter, down, enter, tab, enter. Now I'm back at exactly the same error screen because apparently the install didn't work. Now, I know that you need to be running "npx @zed-industries/claude-code-acp", so I check the docs and apparently I can "toad run COMMAND". But it doesn't work for multi-word commands. And my trial with toad comes to an end.
So I can't test it for anything actually useful right now, but I'll add this to my list of projects to watch. Hopefully, being a UX-focused project, the creator actually focuses on the UX and fixes some of these silly decisions.
This has been my experience also, so far. I like a lot about this and want to start using it, but beyond the initial awkwardness of key bindings feeling wrong, it just seems a bit too early for me. For example, most of the agents I tried to get working on my Arch system failed to connect, only Claude Code and Vibe worked. Most worked on MacOS (except Codex, even though it's installed on my system). But I need to be able to set the agent into Bypass Permissions mode, and when I do, I'm still constantly prompted with permissions checks. There also seem to be weird errors caused by fish shell. I'd also really like to be able to define my own custom agents (eg one use case I'd like is to be able to launch Claude Code but swapping out the Anthropic endpoint for OpenRouter's so I can try new models using CC's agent harness).
It's possible this is just part of the learning curve, but it is making me think I'll have to come back to this project in a month or two to see if there are fewer pain points. Great work so far though.
Pretty sure you are a troll, but I'll give you the benefit of the doubt, Mr cranky pants.
Toad works the same way as a browser. Tab and shift+tab to move focus. Cursor keys move within the control. That's what is happening on the front page. If you don't like the keyboard control, you could always just use your mouse to click stuff.
The agent modal works in the same way. Tab to focus a control. The currently focused control is highlighted with a really obvious accent color. Seriously, this is how web pages work. Try it with this one.
Tiny modal? It had some padding around it, but there is plenty of space for the contents. 100x20 characters IIRC. Unless you have shrunk your terminal down to less than that.
The commands are provided by the creators of the agents. If they fail for any reason, there is probably some kind of error message you could use for tech support. You would get the same result running the command outside of Toad. But I don't think you were interested in fixing it, as you would have mentioned an error message.
> so the OBVIOUS ACTION of RUN THIS AGENT requires pressing tab, enter, down, down, down, enter, tab, enter.
Even more obvious would be to press space. Which is displayed prominently in the footer. How did you miss that? It even works from the front page. Highlight the agent, and press space. No need to open the agent modal.
> But it doesn't work for multi-word commands. And my trial with toad comes to an end.
It does work for multi-word commands. But if your command contains a space you will need to wrap it in quotes. This isn't a Toad thing, this is a CLI thing. I imagine you don't work in the terminal much?
> Pretty sure you are a troll < OK, I'll come clean: I knew what Textualize is before I even evaluated toad, and have never to date found a Textualize app that "sparked joy" as it were. But I figured I should try Toad anyways since A) I strongly resonate with the problem statement of Toad and B) surely the creator of the library would create a best-in-class implementation using that library. Also, https://news.ycombinator.com/item?id=45912796
> If you don't like the keyboard control, you could always just use your mouse to click stuff. < You're reading me backwards. I like keyboard control, but feel like Textualize is a mouse-first UI library. Pressing arrow keys when launching Toad does nothing because I'm not in a "control". As a user, I am supposed to intuitively know that "Recommended" and "Coding agents" are "different controls" and so it should be obvious to me that the arrow keys would not navigate between them?
> 100x20 characters IIRC < You have 20 lines to work with and you decided to shove 4 of the 5 actions for the form into a modal control within the modal dialog. That's my point.
> Which is displayed prominently in the footer. < That's rich. You dedicate 3 lines to the "show in launcher" button and a "Go" button and say the one line at the bottom of the screen is "prominently". Also, that message isn't even on the opening screen of the app, presumably because I have a flawed understanding of what a "control" is.
> I don't think you were interested in fixing it, as you would have mentioned an error message.
No, the error message was perfectly clear! That's how I knew the solution. "Agent returned a failure code: 127 - /bin/sh: claude-code-acp: command not found" The solution is to run via npx.
> It does work for multi-word commands. But if your command contains a space you will need to wrap it in quotes. < You are wrong.
$ toad run "npx @zed-industries/claude-code-acp"
Not a directory: npx @zed-industries/claude-code-acp
> I imagine you don't work in the terminal much? < Please don't blame your users for uncovering bugs that your coding agent put in that you didn't catch in review.
Anyways, go take your sabbatical, relax, you need it. I do genuinely want a better CLI interface for my coding agents and what you said elsewhere that the landscape is "like building a browser for a single website" is very true. Hopefully you come back with fresh eyes and can make a compelling offering in this space.
Very excited to see this come out - though coding agents are impressive their UIs are a bit of a mixed bag.
Textual offers incredibly impressive terminal experiences so I'm very much looking forward to this.
I wonder how much agentic magic it'll be able to include though - Claude Code often seems like a lot of its intelligence comes from the scaffolding, not just the LLM. I'm excited to see!
Hope you like it. It is still Claude Code doing the work. Toad talks to the agent, and is the agent that works with the LLM. So the results should be identical to the native CLI.
I have written a coding agent which I plan to open up soon. By far the biggest time sink has been in the TUI - I've just implemented ACP and I really hope that I can use toad as a front end.
Very much still around... and one of those "this (the terminal tool the HN post is about) might have some trademark challenges" given that every license of TOAD (Tool for Oracle Development) is bundled with some AI and the likelihood of confusion and overlap is quite real. ... and Quest really likes putting that ® after every mention of it.
Check out vibecommander - it’s a young tool in this space with a different take that wraps around CLI coding assistants with IDE-style file and git panels that compliment the experience by letting the human do the code review part of the task seamlessly.
Will add Toad support ASAP, I’m sure they’ll be great together.
You could also consider checking out vim, which solved this 50yrs ago ;) in all seriousness, you have terminals, splits, lazygit… so to me, it seems like this is a case of “not invented here”
Vim has a terrible user experience though. There's a reason everyone stopped using it as soon as they possibly could and moved to other text editors. Now the only vim users are the 60 year old+ greybeards who try to convince everyone they're such morons for not using it.
Stop trying to convince people to use vim, it sucks, it's got a terrible ux, it's not intuitive, it's overly complicated, hard to learn, arcane, and looks like ass.
Toad looks really nice, I will definitely try it out. I have some ACP questions if you don't mind.
First, from my reading of the ACP doc, one thing that seems pretty janky is if the ACP client wants to expose a tool to the agent, e.g. if Toad wanted to add the ability for the agent to display pretty diffs. In the doc they recommend stdio to the ACP server, then stdio to an MCP server, and then some out of band network request back to the ACP client. Have you thought about this, or found a better solution working on Toad?
Similarly, it would be useful to be able to expose a tool which runs a subagent using ACP using a different agent, e.g. if I'm using Claude for coding but I'd like to invoke codex for code review. Have you thought about doing anything like this? Is it feasible over the protocol?
I don’t follow your first question. Toad already displays pretty diffs. MCP works in the same way as the native CLI.
One of the advantages of Toad is that it is vendor agnostic. In the future Toad will be able to run sub agents, and allocate any agent to any job. Still to figure out the UX for that.
1. How has it been working with ACP? Is it anywhere near feature parity with Claude code’s native interface?
2. I see your repo is written in Python which is interesting to me for a responsive TUI. Is it snappy and performant and if so what gave you done to make it feel native? And why did you choose Python?
ACP is will designed. It will always be a few features behind the native CLIs as the protocol catches up. But there is very little that you can't do with ACP. A lot can be done with slash commands that are passed through to the agent verbatim.
Python is more than capable of running a TUI. It is just text manipulation after all. Toad uses Textual, which is currently the best TUI library around. I may be biased saying that as I built it...
I was about to try opencode after using claude code for quite a while.
I think understand the fundamental difference in how they work (acp against existing agentic loops with toad vs a single agentic loop for all models with opencode) but I’m curious why we might want toad over something like opencode, which lets me use any model under the sun.
I suppose toad gets to use the highly specialized agentic loops for each cli. And has a nicer (? opencode is pretty slick from my brief usage…).
Curious to hear about why you chose to built this way and what advantages you see.
It’s stored statically in the Codebase. In the future, I suspect there will be enough compatible agents that there might be a web service to search them.
I think they are working in the Copilot ACP layer. Doubt it will take long.
Sorry, not a question, just wanted to say congrats on putting this together. I am so the target market for a nice terminal interface. I can’t wait to try this out!
The author is also the creator of the textual Python library for creating TUIs. The performance benefits of Rust don't seem very useful in a tool where you spend a few seconds typing in a prompt and then 90% of your time is spent waiting. As long as the UI is responsive when typing there wouldn't be much of a difference.
The creator of Toad, made a TUI framework in Python (Textual). What is so special about Rust, aside from it being blazingly fast and compiled, that you want from it?
Readit News
I fired it up, and the first thing I notice is that the arrow keys don't work. I can't select Claude Code. Oh, apparently it's in a different control, so I have to press Tab, and then the arrow keys work. Wow, this list of buttons has a slow scrolling animation when navigating it. Can I turn that off? Press enter on Claude, now I'm in a tiny modal window. Press enter, because I want to do the obvious thing, but apparently the obvious thing is "show in launcher", so the background of the modal is weirdly changing while a tiny single character inside the button is indicating that this is the part I'm supposed to be focusing on. No, I want to do the obvious thing of running Claude code. You could easily fit the 4 actions of this form on my screen, but by choosing to use a tiny modal window you're now forcing yourself to use another modal drop-down control to choose the action and a separate "yes actually do it" button, so the OBVIOUS ACTION of RUN THIS AGENT requires pressing tab, enter, down, down, down, enter, tab, enter. Great. Now I'm at a chat interface with an error screen, because it isn't installed. Quit the program, restart, enter, tab, enter, down, down, enter, tab, enter to install. It shows a successful run of the "ACP adapter" for claude. Shift-tab, enter, down, enter, tab, enter. Now I'm back at exactly the same error screen because apparently the install didn't work. Now, I know that you need to be running "npx @zed-industries/claude-code-acp", so I check the docs and apparently I can "toad run COMMAND". But it doesn't work for multi-word commands. And my trial with toad comes to an end.
So I can't test it for anything actually useful right now, but I'll add this to my list of projects to watch. Hopefully, being a UX-focused project, the creator actually focuses on the UX and fixes some of these silly decisions.
It's possible this is just part of the learning curve, but it is making me think I'll have to come back to this project in a month or two to see if there are fewer pain points. Great work so far though.
Toad works the same way as a browser. Tab and shift+tab to move focus. Cursor keys move within the control. That's what is happening on the front page. If you don't like the keyboard control, you could always just use your mouse to click stuff.
The agent modal works in the same way. Tab to focus a control. The currently focused control is highlighted with a really obvious accent color. Seriously, this is how web pages work. Try it with this one.
Tiny modal? It had some padding around it, but there is plenty of space for the contents. 100x20 characters IIRC. Unless you have shrunk your terminal down to less than that.
The commands are provided by the creators of the agents. If they fail for any reason, there is probably some kind of error message you could use for tech support. You would get the same result running the command outside of Toad. But I don't think you were interested in fixing it, as you would have mentioned an error message.
> so the OBVIOUS ACTION of RUN THIS AGENT requires pressing tab, enter, down, down, down, enter, tab, enter.
Even more obvious would be to press space. Which is displayed prominently in the footer. How did you miss that? It even works from the front page. Highlight the agent, and press space. No need to open the agent modal.
> But it doesn't work for multi-word commands. And my trial with toad comes to an end.
It does work for multi-word commands. But if your command contains a space you will need to wrap it in quotes. This isn't a Toad thing, this is a CLI thing. I imagine you don't work in the terminal much?
> If you don't like the keyboard control, you could always just use your mouse to click stuff. < You're reading me backwards. I like keyboard control, but feel like Textualize is a mouse-first UI library. Pressing arrow keys when launching Toad does nothing because I'm not in a "control". As a user, I am supposed to intuitively know that "Recommended" and "Coding agents" are "different controls" and so it should be obvious to me that the arrow keys would not navigate between them?
> 100x20 characters IIRC < You have 20 lines to work with and you decided to shove 4 of the 5 actions for the form into a modal control within the modal dialog. That's my point.
> Which is displayed prominently in the footer. < That's rich. You dedicate 3 lines to the "show in launcher" button and a "Go" button and say the one line at the bottom of the screen is "prominently". Also, that message isn't even on the opening screen of the app, presumably because I have a flawed understanding of what a "control" is.
> I don't think you were interested in fixing it, as you would have mentioned an error message.
No, the error message was perfectly clear! That's how I knew the solution. "Agent returned a failure code: 127 - /bin/sh: claude-code-acp: command not found" The solution is to run via npx.
> It does work for multi-word commands. But if your command contains a space you will need to wrap it in quotes. < You are wrong.
> I imagine you don't work in the terminal much? < Please don't blame your users for uncovering bugs that your coding agent put in that you didn't catch in review.Anyways, go take your sabbatical, relax, you need it. I do genuinely want a better CLI interface for my coding agents and what you said elsewhere that the landscape is "like building a browser for a single website" is very true. Hopefully you come back with fresh eyes and can make a compelling offering in this space.
Textual offers incredibly impressive terminal experiences so I'm very much looking forward to this.
I wonder how much agentic magic it'll be able to include though - Claude Code often seems like a lot of its intelligence comes from the scaffolding, not just the LLM. I'm excited to see!
Reading the article I felt like the "sickos: yes!" meme.
Check out vibecommander - it’s a young tool in this space with a different take that wraps around CLI coding assistants with IDE-style file and git panels that compliment the experience by letting the human do the code review part of the task seamlessly.
Will add Toad support ASAP, I’m sure they’ll be great together.
https://github.com/AvitalTamir/vibecommander
Stop trying to convince people to use vim, it sucks, it's got a terrible ux, it's not intuitive, it's overly complicated, hard to learn, arcane, and looks like ass.
First, from my reading of the ACP doc, one thing that seems pretty janky is if the ACP client wants to expose a tool to the agent, e.g. if Toad wanted to add the ability for the agent to display pretty diffs. In the doc they recommend stdio to the ACP server, then stdio to an MCP server, and then some out of band network request back to the ACP client. Have you thought about this, or found a better solution working on Toad?
Similarly, it would be useful to be able to expose a tool which runs a subagent using ACP using a different agent, e.g. if I'm using Claude for coding but I'd like to invoke codex for code review. Have you thought about doing anything like this? Is it feasible over the protocol?
One of the advantages of Toad is that it is vendor agnostic. In the future Toad will be able to run sub agents, and allocate any agent to any job. Still to figure out the UX for that.
1. How has it been working with ACP? Is it anywhere near feature parity with Claude code’s native interface?
2. I see your repo is written in Python which is interesting to me for a responsive TUI. Is it snappy and performant and if so what gave you done to make it feel native? And why did you choose Python?
Python is more than capable of running a TUI. It is just text manipulation after all. Toad uses Textual, which is currently the best TUI library around. I may be biased saying that as I built it...
I was about to try opencode after using claude code for quite a while.
I think understand the fundamental difference in how they work (acp against existing agentic loops with toad vs a single agentic loop for all models with opencode) but I’m curious why we might want toad over something like opencode, which lets me use any model under the sun.
I suppose toad gets to use the highly specialized agentic loops for each cli. And has a nicer (? opencode is pretty slick from my brief usage…).
Curious to hear about why you chose to built this way and what advantages you see.
You get a nice UI that is only going to get better as time goes on.
It’s far better model to separate the agent from the UI. The current situation is like building a browser for a single website.
How are new agents added? Do you have to write a dedicated plugin for each one? Or there's some kind of discovery mechanism?
(I was looking for Copilot, but I guess that will depend on https://github.com/github/copilot-cli/issues/222 ?)
I think they are working in the Copilot ACP layer. Doubt it will take long.
And Copilot isn't supported for now because, well, there is no ACP support
Ideally working with toad to experiment with it.
If an agent can be configured to use Ollama, then you could use it from a Toad. It might be possible right now.