There is an outstanding feature request around it at least:
I think it's another workflow, that might suit your needs, but I think the real magic continues to be in the model. Similarly, I'm sure Claudia adds some nice window dressing.
Eventually folks will settle in on some local maxima for interaction and software development with LLMs. Who know what it will look like? It'd be nice if whatever comes next bumps the industry out of the current scrum and sprint style workflows, if only to break folks out of the theatrics of these software personas and rituals.
Here's my WIP: https://github.com/astral-drama/filter
1. https://github.com/anthropics/devcontainer-features 2. shameless plug, my own open-source CLI for doing this: https://github.com/smithclay/claudetainer
Step 2: Turn around and sell security-as-a-service to the most profitable products
I am not saying it's infringement, I am just saying that my dumb brain made that connection and I feel like it's not unreasonable to assume that other people might as well.
For once reading the comments first has paid off!
Same reason you can't release a handheld console called the Gamegirl, or a voice assistant called Alexis.
I wonder if I'll have to rename at some point.
As others have said, this is a giant red flag.
It already worries me that the Cursor agents occasionally try to perform operations with full absolute paths, which they wouldn't be able to know if they were properly sandboxed to the current directory.
On the other, the trend seems to be everyone developing a million disparate tools that largely replicate the same functionality with the primary variation being greater-or-lesser lock-in to a particular set of services.
This is about the third tool this week I've taken a quick look at and thought "I don't see what this offers me that I don't already have with Roo, except only using Claude."
We're going to have to hit a collapse and consolidation cycle eventually, here. There's absolutely room for multiple options to thrive, but most of what I've seen lately has been "reimplement more or less the same thing in a slightly different wrapper."
Because I mentioned it and it's what I use daily: Roo is a VSCode extension. So you get the entire VSCode ecosystem for free. On the AI specific side, it has every feature this app highlights on its homepage and more. It works with just about any API provider and model you could ask for.
I could probably translate my existing workflow over to Claudia pretty easily, but what does that get me? A slightly different interface seems to be about it.
That's the question I keep hitting with these new tool announcements.
Continue.dev has some features, but it’s on VSCode and Jetbrains
you give up one side of freedom (the ide) for the other (the backend).
That said, VSCode is a popular platform for this for exactly the reason I think consolidation is eventually inevitable: it's got a huge preexisting ecosystem. There are extensions for practically anything you could ask for.
There's likely room for some standalone, focused apps in this space. I just don't see the current wave of "we put a wrapper around Claude Code and gave it some basic MCP and custom prompt management tools like a dozen other applications this week" being sustainable.
They're all going to end up on their own tiny islands unless there's a reason for an ecosystem to develop around them.
A tiny island is fine for a tool like this - not everything needs an 'ecosystem'.
Typically, applications develop a userbase when they offer something that people can't find elsewhere.
What I'm saying isn't "everyone should be using VScode extensions for this"; it's "I see nothing to distinguish this from a bunch of other functionally identical applications and people just keep building them." I literally don't see a single unique feature promoted on the landing page.
My fundamental point is that we're in a gold rush phase where people are all building the same thing. We'll eventually see a handful of apps get popular and effort swell around those instead of everyone reimplementing the same thing. And my money is on that looking a lot like it usually does: the winners will be the apps that find some way to differentiate themselves.
Which is super cool. Like during the dawn of web 2.0 we had lots of aggregators and forums instead of "Reddit and others."
(I'm not saying it's good UX.)
My complaint is more that right now it feels like everybody is rushing to fill the exact same space with the exact same feature sets.
It's resulting in a lot of superficial diversity that's functionally homogenous. I want to see more applications that are pushing the capabilities of current AI tooling in creative directions.
Otherwise, you're going to see the variations on the same thing over and over, which is totally fine, and where innovation comes from.
Personally, I just use stock VS Code (copilot) and Cursor.
So, in other words, this is the exact opposite? “Lost of aggregators and forums” meant diversity. Lots of small players doing their own thing. What we have now is a handful of big players, and then tons of small players accessing those services with a different coat of paint. It’s like if the web you mention consisted of lots of people doing alternative interfaces to access Facebook and Reddit.
So… what we had on web 2 then, with its daily twitter clients? There were hundreds and hundreds of them
But long story short she showed me what she had on her iphone and it was a totally different app that wrapped a text chat interface around chatgpt, it wasn’t even themed like to be a persona or anything but was at the expense of any multimodal capabilities
Just caught me off guard about how common that might be
Seems pretty scammy to me, akin to typo squatting with potential to collect a lot more personal information but he can’t always be reasoned with.
Hopefully he heeds my advice to not provide anything personal.
> I don't see what this offers me that I don't already have with Roo
Ironic >-< for an AI tool tied to a specific IDE
How many ways can you wrap (multiple agents, worktrees, file manager, diff viewer, accept reject loops, preset specifications for agents) -- let's try Electron! Let's try Tauri! Let's try a different TUI!
What if we sat down and really thought about how these agentic IDEs should feel first instead of copy pasting the ideas to get something out to acquire market and mind share? That's significantly harder, and more worthwhile.
That's how these agentic front ends should be advertised: "Claude Code, plus _our special feature_" and then one can immediately see if the software is filled or devoid of interesting ideas.
Wrappers constantly live in the support and feature parity of today.
Anthropic’s Claude Code will look a hell of a lot different a year from now, probably more like an OS for developers and Claude Agent non-tech. Regardless they are eating the stack.
Pricing/usage will be very simple - a fixed subscription and we will no longer know the tokenomics because the provider will have greatly abstracted and optimized the cost per token, favoring a model that they can optimize margin against a fixed revenue floor.
This is definitely not how most compute-constrained cloud services end up looking. Your cloud storage provider doesn't charge you a flat rate for 5tb/month of storage, and no amount of financier economics can get Claude there either.
Training is capital-intensive, yes, but so far it appears that there will always be some entities willing to train models and release them for free. All it takes is a slowdown at the frontier for the open models to catch up.
The money is in the hardware, not the software.
Qwen Coder 30B is my main driver in this configuration and in my experience is quite capable. It runs at 80 tok/s on my M3 Max and I'm able to use it for about 30-50% of my coding tasks, the most menial ones. I am exploring ways to RL its approach to coding so it fits my style a bit more and it's a very exciting prospect whenever I manage to figure it out.
The missing link is autocomplete since Roo only solves the agent part. Continue.dev does a decent job at that but you really want to pair it with a high performance, large context model (so it fits multiple code sections + your recent changes + context about the repo and gives fast suggestions) and that doesn't seem feasible or enjoyable yet in a fully local setup.
When I use qwen coder 30B directly to create a small demo web page, it gives me apl the files and filenames. When I do the same thing in roo chat (set to coder) and it runs around in circles, doesn't build multiple files and eventually crashes out.
The main problem I'm seeing, is that a lot of the tooling doesn't work as well "agentically" with the models. (Most of these tools say something like 'works best with Claude, tested with Claude, good luck with any local models'). The local models via LM Studio already works really well for pure chat, but occasionally trip up semi-regularly on basic things, like writing files or running commands -- stuff that say, GitHub Copilot has mostly already polished.
But those are basically just bugs in tooling that will likely get fixed. The local-only setup is behind the current commercial market -- but not much behind.
I strongly agree with the commenter above, if the commercial models and tooling slow down at any point, the free/open models and tooling will absolutely catch up -- I'd guess within 9 months or so.
Personally, I think it's far more likely that a year from now either SotA models will have shifted elsewhere or Anthropic will have changed their pricing model to something less favorable than the current MAX plans. Either of those scenarios could suddenly result in the current Claude subscription models either not existing or no longer being the screaming deal they are now. I think it's exceedingly unlikely we see any major provider go to an unmetered business model any time soon.
And if you've built your entire workflow around tooling specific to Anthropic's services, suddenly you have an even bigger problem than just switching to a more cost effective provider. That's one of the bigger reasons I'm very skeptical of these wrappers around CC generally.
Even Claude Code itself isn't doing anything that couldn't and hasn't been done by other tools other than being tied to a really cheap way to use Claude.
Even on a Max plan, it's not hard to completely blow through your usage limits if you try to use Opus heavily.
All it takes is another provider to land a combination of model and cost that makes Code less of a deal for vendor lock-in to become a problem.
So depending on the parent company, they may prefer to have a - to be a little enterprisey - set of ISVs that are better in specifc domains.
- Happy Claude Code Client: open source (MIT) effort for a quality mobile app
- Omnara: closed source mobile app, $9/month
- CodeRemote: closed source mobile app, $49/month
- Kisuke: closed source mobile app, private beta, unknown price
If you know of others, I would appreciate a PR to update the table I put together, or just let me know and I'll add it.
https://happy.engineering/docs/comparisons/alternatives/#qui...
There are more more desktop apps, probably because those are easier to design.
I’ve started using conductor.build and it feels nice, but would happily evaluate others.
If there's no particular feature that only Claude offers, this is just needless vendor lock-in. And what happens if another lab releases a model that suddenly trounces Claude at coding? Your users will leave for an app that supports the new hotness, and you won't be able to keep them because of a short-sighted architecture that cannot swap model providers.
I mostly use Claude Code with a Max plan via Roo. I have the option of sending prompts to OpenRouter if I've hit usage limits or if I want to try a particular task with a different model (e.g., I'll sometimes flip to Gemini Pro if a particular task could benefit its large context windows).
Besides that. These tools are changing so fast that to build an agent agnostic tool would be insane given the speed and market pressures right now. Why support roo or cline or cursor cli if it adds 3x engineering cost for 20% more market reach? The reality is there are no standards around the way the actual leading tools work if you wanna build something on Claude/codex/(insert flavor of the week).
Gotta pick your horse and try to hang on, and hope you picked right.
What a ridiculous proposition. It’s me making the app right? You getting to use it (if you want) is purely incidental. If you never use it because it doesn’t support anything but claude, that’s not something I consider a problem.
I built some workflows using Claude’s API and now wish I had used a wrapper so I could easily switch to try gpt-5 for the cost savings.
1. Real-time sync of CLI coding agent state to your phone. Granted this doesn't give you any new coding capabilities, you won't be making any different changes from your phone. And I would still chose to make a code change on my computer. But the fact that it's only slightly worse (you just wish you had a bigger screen) is still an innovation. Making Claude Code usable from anywhere changes when you can work, even if it doesn't change what you can do. I wrote a post trying to explain why this matters in practice. https://happy.engineering/docs/features/real-time-sync/
2. Another contributor is experimenting with a separate voice agent in between you and Claude Code. I've found it usable and maybe even nice? The voice agent acts like a buffer to collect and compact half backed think out loud ideas into slightly better commands for Claude Code. Another contributor wrote a blog post about why voice coding on your phone while out of the house is useful. They explained it better than I can. https://happy.engineering/docs/features/voice-coding-with-cl...
The only thing I need this point is a push notification when it needs an approval or it has stopped.
https://happy.engineering/docs/guides/self-hosting/
https://github.com/slopus/happy-server
There is a Dockerfile, plus a docker-compose.yaml, and a complete Kubernetes template as well.
Plus I wanted a system that did not require the app and the computer running the AI agent to both be online at the same time. Having a third computer act as a dumb mailbox handles some corner cases I care about.
I've been trying to surreptitiously get claude code and an oven specific MCP server to run on my friend's smart oven for a prank. However this oven enters a low power state when you don't interact with it; killing the network connection. My vision is to queue up commands via the mobile app with fuzzy logic, and then have the oven make weird noises as determined by claude code at some later point when they go to make a pizza or something.
There are already a plugins to use claude code in other IDEs.
This “Ill write a whole IDE because you get the best UX” seems like its a bit of a fallacy.
There are lots of ways you could do that.
A standalone application is just convenient for your business/startup/cross sell/whatever.
All the more reason to embrace a fully open source stack. We need to go hard on "lesser".
Lots of competing actors doing lots of similar things with confusing comparisons and quantifiable results.
By the way, I did not wait for the Claudia demo to load, I was on the website for like 10 seconds, still did not load so... okay then.
Don't worry, someday you'll grow wiser and maybe you'll spend some time living as a CLI executable.
Yes, please :)
At most, I've been thinking about installing one of the extensions to integrate Claude Code into (neo)vim, but even that I'm not sure I really want or need.
But for people who arm themselves to the teeth with GUIs and IDEs, I guess I can see the appeal.
What I want at the core is to be able to open up access to my laptop's currently running Claude Code instance (without all these hacky backdoors that fork the chat with every message by using `--print`; I want a first class API that lets me append messages to the current chat), then I want to be able to send messages (with voice transcription) and approve/deny permissions and see the code diffs and all of that.
Maybe something like a Telegram bot? I had hopes for Claude Code UI[1] but the web interface is too clunky on mobile.
https://github.com/amantus-ai/vibetunnel
Many of my claude tasks get frequently get stuck over small stuff, or require input from me, so I had Claude whip up something comparable in a few minutes. Now I can keep them moving remotely with ease and not be stuck in front of a desktop monitoring them! It's dreamy.
(Not to mention that if I only have my phone, I'm probably out doing something where I don't want to be working...)
My phone also can build and run many projects on its own so I often don't even connect to the laptop.
I’d really love to know what specifically about this is beautiful.
Steve Jobs says it and for 20+ years every tech bro parrots it mindlessly. “Oh do I think I did a good job on this? It must be beautifully designed.”
LLMs are that, but claude code is not.
I see a lot of commenters asking why a GUI is necessary. When you're running several agents in parallel it becomes very handy compared to the terminal. I can easily see the status of each which I haven't found a good equivalent for when using terminal tabs. Also it handles automatically creating git worktrees for each agent which is great.
You can tile terminals, you can use things like tmux to insert multiple command lines into one window, etc.
Question about Dashboard - I'm on subscription plan, and Dashboard shows - "Total Cost: $XX.YY". Is it somehow representitive towards how much credit I used in my plan, or it's just showing a costs it would take if I was using API instead?
What people do care about is where names imply some relationship that isn't true. In this case the name Claudia strongly implies that this is an official Claude-related product. They'll get a cease and desist soon enough if this actually becomes popular.
Thanks, OP!
But yeah - I need to bounce between multiple OS so I avoid tools that don't run on all major desktop platforms.
Also saw this a couple months ago.
like a precursor to reddit's own API pricing changes that made it hard for 3rd party clients to compete.
The saving grace with these API wrappers is that local models being a thing can still let them hedge against the underlying AI labs eating up their stack.
Tweetdeck came out as the leader over the rest (mainly due to having actual functionality), and about a year after it was the clear victor (2011) twitter acquired them, and slowly integrated them into twitter properly (though they killed a bunch of features along the way intentionally).
Fast forward to the musk takeover, and twitter's API pricing changed such that making a third-party client is infeasible.
I think a lot of the same is likely to apply here.
It's a winner-takes-all market, there's a bunch of people iterating on form and ignoring function, and the winner will be based more on function than form.
If there's a clear winner on function, one of the AI companies can acquire and integrate it.
In a way this is probably the future state - 1000 different clients for 1000 different people, each fully customized to their taste
Still miss Apollo
-----
Magnet - The AI workspace for agentic coding
------
You can think of Magnet as your workspace for collaborating with your human & AI agent team mates. We let you quickly spin up Claude Code sandboxes for every issue, and we're also thinking about how AI can be more of a thought partner in building high-quality software.
We're thinking about this problem space more broadly than just trying to be a GUI for Claude Code (though that's already a great starting point).
These are a few of the themes we think about:
- How can we use AI to help you think critically about the features you're prioritizing and what to build next?
- How can we always assemble and provide exactly the right context for every issue you're working on?
- What are the best patterns for collaborating with your human & AI teammates, to ship the highest quality code?
- How can you best specify exactly what you want, and verify that it's what you hoped for?
Would love for y'all to try it, and I'll post a video of me building a product with Magnet a little later here - the tool's getting really fun to use!
We're also very open to feedback and try to incorporate learnings quickly! I spent a large part of this weekend using Magnet to fix most of the issues someone we onboarded Friday brought up
It wasn't great.
- Installation using the provided binaries just fails on my machine - I have Ubuntu 22.04, which apparently has too old a version of glibc. Building from sourced worked though.
- Every time I want to open a new chat, it brings me back to the project list. I don't want to click on the same project every time!
- Scrolling is awful! It's slow, and it often doesn't automatically scroll down as the chat is generated so you have to do it yourself.
- There's no title or anything across sessions. If I'm now working on multiple things at the same time, I want to know what I'm working on quickly!
- The log/text entries take up so much space. Something like this would benefit from a much more compact view - it shouldn't use my entire screen to show me 1 TODO list and 1 tool use.
- Unlike the video, the code changes are all wrapped in a "AI Summary" entry which tells me what it did in a few words, with no option (that I could find) to open the code itself. Confused, couldn't find a setting for this.
- There's multiple UI bugs, and it's sluggish overall.
I didn't use the Agents stuff, which (given the video starts with it) might be the main focus? But as it stands, for my attempt at running multiple Claude Code sessions at once, this was too buggy to really work. Someone else mentioned https://conductor.build/, which might be more what I'm looking for, but unfortunately it lacks Linux support.
I hope it gets better! I could see myself using it after a few more releases, and I'm rooting for them - just sharing my experience here for others who are considering trying it.
I'm not as tied to the cli as other folks here, but even I found the Claude Code cli to be a better experience than this too.
I think it will improve, but for now I'm sticking with the cli.
I don't want to install a tool like this that isn't made by Anthropic.
I, too, find this confusing and actively dislike the naming and association here.
When I want to start an agent, I run a small command that takes in a prompt (and some optional flags but mostly the defaults are what I want). It does this: 1. Check out a worktree from a repo (defaults to my bread and butter repo with completions for the most active ~5 repos in my org) 2. Craft a command to launch an agent (supports codex, opencode, and gemini-cli because my company only gives us OpenAI keys). 3. Launch the command in a tmux session. 4. Exit 0.
There are some light integrations with Jira, GitHub, etc but they're accomplished by shelling out to other tools.
If I want to manage sessions, I've got tmux + fzf + other unixy tools old & new to manage them.
If I want to manage sessions remotely, there is vibetunnel. It would be cool to have a slicker AFK experience, but it works.
My silly little piece of vibe coded slop and duct tape, for my use case, is largely competitive with most of the offerings on the market, outside of the ones that do cloud-based environments. Some of these projects are VC funded teams of people working full time, I'm assuming. What a time to be alive.
I'm also amazed at how readily projects like this just embrace Claude Code lock-in. Is there really anything specific about it that other agent harnesses don't support? I haven't used it yet, but so far it just seems like it benefits in mindshare alone from being the default/first mover, not because it supports any particular feature the others do not. TBF I do hear that it is quite good.
Quality & polish is a compelling reason to _use_ something, but it's not a compelling reason to build walls IMO, especially in a space like this where context engineering techniques, prompts, etc are in no way secret sauce and can be readily copied.
CC development is not just development, not all types of development. it’s frontend JS based , and it’s backend development. Only those scenarios work.
Try creating native desktop or mobile app and it’s like a swamp of trial and error.
You have to learn by trial and error what documentation sets and instructions you have to provide at which moment and context and balance at with token cost.. it’s a multi dimensional problem for which there are no recipes that work.
On top of that your direct instructions to not use particular patterns or approaches gets forgotten and ignored y CC with later “you’re right, I should have…”. I am starting to think it’s not solvable by the user by providing docs, examples and instructions. That Claude must have native development baked in to the same level as they baked in the frontend and backend.
What I am getting to is - make a tool to manage those doc sets and contexts and instructions and allow to share those sets between users globally as recipes.
> Try creating native desktop or mobile app and it’s like a swamp of trial and error.
I had it build an Android app and embedded code for a BLE peripheral at the same time, tailing the logs from each to debug issues. It worked great.
I generally avoid stating in concrete terms what Claude Code “can’t” do. Especially since it keeps getting new built-in features (like the ability to background shell processes and then tail their logs later) that often fix issues people talk about having had a month ago.
I’ve noticed these UIs tend to strip away native features (plan mode, sub-agents etc.). A thin layer that preserves Claude Code’s built-ins while orchestrating instances, worktrees, and branches from the shell would be ideal for me.
* IR: https://register.dpma.de/DPMAregister/marke/registerIR?AKZ=1...
* EU: https://register.dpma.de/DPMAregister/marke/registerHABM?AKZ...
Edit: And the "getAsterisk" organization at https://github.com/getAsterisk/claudia and https://asterisk.so/ which is NOT about https://www.asterisk.org/. PSA: The real Asterisk has class 42: https://register.dpma.de/DPMAregister/marke/registerIR?AKZ=9... - Oh boy. What next? Call the next project TheRealMircosoftGoogle? Why not? Lol.
`Error: Problem parsing d="M 16.66% 50% L 83.33% 50%"`
It works very fast compared to Cursor -- biggest limitation is that it doesn't have workspaces where you can get context from several repos for building your application.
MCP servers also work as well just a little more complex.
Then, when I create a new one, I tell Warp to use this folder as the boilerplate for the new one. It works perfectly.
Go play with it, but don't use for important chats already.
sibeliuss•5mo ago
(Super rapid zooming in and out, flying all over screen at 3x speed, must cover eyes!!)
nperez•5mo ago
Tempest1981•5mo ago
Extreme close-up: https://www.youtube.com/watch?v=YdxsWw_gV3E
canogat•5mo ago
svantana•5mo ago
dcreater•5mo ago
Made something? Dont start with a long, protracted "This is how i built it from step 1". Show the damn thing! Then tell me how you built which i will be interested in only if the product is good to begin with.
fullstackwife•5mo ago