The Agent Model came out very recently, I’ve been following the GitHub issue over the past days and you can see it was rushed out. But I don’t see anything wrong with that, many AI topics are being rushed out and adding slash commands and other small things are very small things to add once the foundation is there.

Tbf I never use /compact but clear instead, and load in the relevant context anew. I just haven’t seen compacted context to be very useful, so far.

Seems like those issues are largely limited by SDK so urging Anthropic to adopt is the only realistic move

Yeah I was initially excited here, but it feels more like a demonstration of what's possible rather than a working tool.

I found the interface very nice but quickly ran up against limitations on prompt length (it wasn't that long) for example. I am used to being able to give detailed instructions, or even paste in errors/tracebacks.

I'll check back in in a few months.

I was somewhat surprised to find that Zed still doesn't have a way to add your own local autocomplete AI using something like Ollama. Something like Qwen 2.5 coder at a tiny 1.5b parameters will work just fine for the stuff that I want. It runs fast and works when I'm between internet connections too.

I'd also like to see a company like Zed allow me to buy a license of their autocomplete AI model to run locally rather than renting and running it on their servers.

I'd also pay for something in the 10-15b parameter range that used more limited training data focused almost entirely on programming documentation and books along with professional business writing. Something with the coding knowledge of Qwen Coder combined with the professionalism and predictability of IBM Granite 3. I'd pay quite a lot for such an agent (especially if it got updates every couple of months that worked in new documentation, bugfixes, github threads, etc to keep the answers up-to-date).

I'll third this. AI autocomplete is THE most efficient and helpful feature of Cursor, not the agents.

What Zed lacks in code generation quality it makes up for in not-being-an-Electron-app

I wonder if Augment [1] are working on a Zed plugin.

I've been using Augment for more than a year in Jetbrains IDEs, and been very impressed by it, both the autocomplete and the Cursor-style agent. I've looked at Cursor and couldn't figure out why anyone needed to use a dedicated IDE when Augment exists as a plugin. Colleagues who have used Cursor have switched to Augment and say it's better.

Seems to me like Augment is an AI tool flying under most people's radar; not sure why it's not all over Hacker News.

[1] https://www.augmentcode.com/

I now plain hate Cursor's auto complete, it's too aggressive I cannot write any code anymore, it seems to have hijacked CMD too, not just tab.

I'd also like to second this and probably will in every Zed post. This is the primary reason I'm not ready to switch to Zed just yet.

>One thing that still suffers is AI autocomplete. While I tried Zed's own solution and supermaven (now part of Cursor), I still find Cursor's AI autocomplete and predictions much more accurate (even pulling up a file via search is more accurate in Cursor).

It's not only the autocomplete. I've never had any issue with Cursor while Zed panicked, crashed and behaved inconsistently often (the login indicator would flicker between states while you were logged in and vice versa, clicking some menus would crash it and similar annoyances). Another strange thing I've observed is the reminder in the UI that rating an AI prompt would send your _entire chat history_ to Zed, which might be a major red flag for many people. One could accidentally rate it without being aware of that and then Zed has access to large and potentially sensitive parts of your company's code - I can't imagine any company being happy with that.

>I am glad to hear that Zed got a round of funding. https://zed.dev/blog/sequoia-backs-zed

There are plenty of great VCs out there, going with Sequoia will definitely come with some unpleasant late consequences.

>This will go a long way to creating real competition to Cursor in the form of a quality IDE not built on VSCode

There are many "real competitors" to Cursor, like Windsurf, (Neo-)Vim, Helix, Emacs, Jetbrains. It's also worth being aware that not everybody is too excited about letting AI slop be the dominant part of their work. Some people prefer sprinkling a little AI here and there, instead of letting it do pretty much everything.

Glad it was helpful :)

I’ll keep an eye on this ‘proper’ Zed support for sure, although the current setup is working just fine so I might wait for v0.2.

"even pulling up a file via search is more accurate in Cursor"

Huh? it takes it sometimes like 40s to find some file with the fuzzy search for me. In that time im going to the terminal running a "find" command with lots of * before I get some result in cursor

I agree, I've fantasized about an editor with a truly pluggable editing model which is decoupled from the other parts.

Yi was kind of designed like this, I believe. You could compile in an emacs-like model, a vim-like model, or presumably make your own model.

I've used Helix and Kakoune in addition to Emacs and Vim, but dealing with the limitations/featureset/plugin treadmill gets a little tiring.

I have been following Zed, and it seems that they have rearchitected things to enable adding Helix mode and making the editing model a bit more modular, but it's still fairly new. They are fixing bugs pretty quickly. I will have to try it again.

They have a nice discussion here:

https://github.com/zed-industries/zed/discussions/6447

They reference Ki, which also looks cool, and they out some of Helix's inconsistencies in their comparison: https://ki-editor.github.io/ki-editor/docs/comparisons/

I prefered Kakoune to Helix (it was more consistent). But to your point, being able to swap these things out more easily would let you choose an editor based on features, and not tradeoff between features and an ergonomic editing model.

Ironically you can use Ki inside of VSCode (and I know you can use Vim that way too), but VSCode is so darn bloated and slow...

It’s exciting that Zed even has a Helix mode. That was a big moment for Helix.

Last time I tried it, though, I immediately ran into parts of the keymap that hadn’t been translated yet. I’m already at my limit of tools in beta mode/built from my own fork, so I switched back to Vim mode – where the team is on record explaining their thorough testing methodology.

As a Helix user of two years, I sometimes wonder if I actually like the Helix keymap (certainly some parts are nicer than Vim’s) or if I simply tolerate it because of how nice it is to get a polished TUI IDE out of the box. Either way, my muscle memory expects Helix mode now, rather than Vim.

Neovim can run in server mode, where other editors send it user input and then Neovim sends back the buffer. This is how I use vim in VSCode — not the Vim extension but the Neovim extension, which uses the real Neovim, which of course reads my Neovim config and plugins and makes them available to VSCode. So it seems like helix “just” needs a server mode, and then you can integrate it into any editor.

Helix seems to have good LSP support from what I can tell? The only language I use at $WORK that doesn't have full support is GraphQL which lacks auto indent.

If you want to try something similar to Helix in emacs, there's meow-mode. While I'm not a helix user myself, it shouldn't be too difficult to get meow to work like helix.

If you are already familiar with Vim bindings is Helix's object then action really worth that much?

Helix itself seems quite young - and first time I’m hearing of it.

I had the exact same problem. I was so stoked to try helix mode and then realized it obviously doesn’t have any of my backspace shortcuts. Duh, but still… back to helix!

Oh what, there’s a Helix mode now?

Been wanting to learn Helix more and using it for small edits but never saw a Helix mode in any editor yet

The nuance situation is rapidly improving. I had several minor issues with Zed ~6 months ago, and most of them have been patched away.

For what it's worth, I think Zed now has a default keybind to open settings.json: Ctrl+,

I assume that keybind is also configurable?

I spent a while trying to set it up, as I share your general take on their ethos. Personally, I'm okay with a 'power user'-focused text editor, even! But the relative lack of syntax highlighting options got me to give up. Maybe I'm just spoiled from SublimeText's dope, complex, extensible system for specifying "contexts" in themes, but Zed was just nowhere near enough for me.

The keybinding system is also nuts if you turn on Vim mode, but I think I'd eventually get used to that. But functions need to be a different color than arguments, which need to be a different color than local variables... Just non-negotiable.

I look forward to trying it again sometime soon! The AI features seem rad, this included.

Zed does have a way to run LLMs without tool calling. From the agent pane, in the menu, select “new text thread”. I believe there’s a keyboard shortcut but I’m on my phone right now.

That's unfortunate. I use Zed and I'm moving towards containerising my dev environment (using SSH remote dev to connect Zed to the container) because all this agentic stuff seems like a security nightmare. At the very least I want to restrict the blast radius to my repos dir.

Remote dev isn't very good in Zed, unfortunately. For some reason they chose not to apply the local editor's settings to a remote session by default. Every remote has its own config file. Questionable choice, imo

Any reason for this? Is it something temporary or it will never be supported?

IIRC it doesn’t work in Cursor either, and their own AI sidebar was getting weird issues too. Mostly switched back to VSCode for SSH workflows because of that.

zed crashes completely now for me when using a remote env over ssh since a couple releases ago.

Really love Zed after working in it full time for a month now and pay their 20$ sub tier to support them even when I rarely use the LLM integration beyond the auto-complete.

At first I was very dismissive of it due to being Apple-first but they've turned it around with really good Linux support and it seems like Windows soon as well!

Zed being OS is 100% what's keeping me from moving away. If anything happens you just fork it and accept it as-is

You basically have to choose from

- Zed: VC funded open source

- Sublime Text: indie closed source

None is ideal, but I guess we all know why.

I use zed and claude code side by side right now. I haven't tried out the newly released assisted agent mode with Zed.

Yes Opus has been good with instruction following and same with Gemini for 2nd opinions and brainstorming.

They're not perfect but definitely I see plenty of value in both tools as far as they are reliable services.

I don't like the cloud based functioning of the models as the experience is extremely flaky and not reliable. I've gound OpenAI Codex and the models in codex too be more reliable in responses and consistency of the quality service.

I would still prefer to have a fully locally hosted equivalent of what ever the state of the art coding assisstant models to speed up work.

That will take time though as in with every technological evolution. We will be stuck with time sharing for sometime haha. Until the resource aspect of this technology scales and economizes to become ubiquitous.

Maybe there's a way to use your Max plan? I heard OpenAI works the same way, there's a way to use the codex cli with your chatgpt subscription (I think it gives it a jwt token or something). There are 3rd party tools that can then use the same token to make API requests via the subscription, instead of via API billing. (Which is sort of allowed, since the cli is open source?)

I'm guessing claude code works roughly the same way?

>h Opus 4.1 for helping me with my most challenging tasks

I switched to gemini 2.5 pro, after some prompt tweaking nothing really beats in actual coding tasks imho.

Honestly, I’d love to know why that happens. Can you take a look at the logs? On Mac, you can do it through the Console app, and on Linux, through idevicesyslog.

Anyone running this on Linux? I find it works fairly poorly there. To be fair, vscode is also not great for me (especially vim mode) on Linux.

What's so great about zed ?

What do you feel is missing from the UI?

> I don't see why most editors should behave like hand crafted musical instruments when in reality they are much more akin to high quality knives in a kitchen (sure you have your favorite knife set and bring it from job to job, but at the end of the day you can be just as productive with a different knife when necessary).

This is such a poor analogy. Yes, a good chef can make do with a different knife, but there is a reason why chefs pay for significantly higher quality knives, keep them sharpened, and treat them with diligence and care, than other kitchen tools. A blunt knife can actually be dangerous. Consequently, a lot of chefs buy knives that are effectively hand crafted / forged knives out of this relentless pursuit of quality.

> What most of these comments are missing is the attempt at standardization and unification.

> While that may be true and your particular workflow requires certain features, what is overlooked is the survival pressure for editors.

I think your general perception is not something I agree with. I want to use software I enjoy using. Programming is a creative exercise for me, and I want to use the tools I enjoy. If a tool is not enjoyable to use, I do not want to use it. Sometimes, productivity does increase enjoyment, but sometimes it doesn't. For example, arguably I would have been more productive in my Java days if I used Eclipse, but because the editor was so bad, I preferred to learn the APIs myself and use Sublime Text instead.

I also don't think I'm sympathetic to the survival of any particular editor. Software comes and goes, and sustainably built business models will prevail. All of the AI-first editors hinge on this being the right iteration of this technology, and we simply do not have a long enough timeframe or context to know if this is truly the best way to write code using AI. MCP/ACP, whatever else might be the best strategy for now, but I think it's too early for anyone to suggest that we've come to the right conclusion forever.

I'm sorry if this is blunt but is Agent Client Protocol... ...good?

It just looks to me like a bolted-on dongle to the past 50 years of kludges in editor design. It hasn't got 1/20th of the value proposition that a proper shared state layer would offer.

You should try Magnet for this! (www.magnet.run) https://www.loom.com/share/ceabf385871e45b88fcb512b6de6b684

I suspect Zed will aim to tackle this issue via DeltaDB:

https://zed.dev/blog/sequoia-backs-zed#introducing-deltadb-o...

Same, logged an issue: https://github.com/zed-industries/claude-code-acp/issues/18

Ona makes this easy: https://ona.com/

It’s also great that they’re willing to risk that, in the name of a potentially better user experience. That’s what gets them to win in the long run, not building another walled garden.

Could be cheaper, maybe?

I have been using zed a fair bit with clause api and the 50 free prompts a month that zed provide on the free plan works out to roughly $8 for an equivalent amount of code using the claude api.

But no idea how it compares to the subscription plan.

It's really interesting point of view. I'm one of those people who avoid using VSCode at any cost. It's slow, it's bloated, the UI is not great, and it's slowly being locked down by Microsoft.

If Zed would not exist, I would be using helix, neovim, or emacs as I did before.

Zed's main selling point over VSCode for me is the lack of a slight delay between when I press a key and when the character appears.

VSCode has always felt ever so slightly sluggish to me, and I find it maddening as I type.

There is always a "better mousetrap", and there are those that continue to use the old one because they "know how it works and it's set up just the way I like it". And there are others that try every new mousetrap that hits the market. (and that's ok, not slighting either one)

I will say that I personally have never really gelled with VSCode no matter how much I try to customize it, it still is just a bit off. For me, it's like it's too much to be a simple editor like SublimeText or NeoVim, but not quite enough to be an IDE like IntelliJ or Visual Studio (full). It does just enough that I expect a bit more of it and it often fails to deliver. Right now I tend to just use 2 editors - one very simple one for viewing/editing text files and one IDE (currently IntelliJ) for coding in a project.

On topic - Zed is actually a really nice editor. It had some rough edges last time I tried it, but it's probably about time to give it another go.

I wanted to like VSCode but it has enough input latency on my machines that it's not that enjoyable when I'm "locked in". Also if I'm running a bunch of services in Docker on MacOS (which means they're running VMs sigh) the overhead of VSCode is just too much and the system starts swapping constantly grinding the whole thing to a halt. I also find configuring it a pain. Every configuration pane feels ad-hoc and not part of a holistic, configurable system. Emacs has lots of crusty bits and an annoying event loop that you have to really work around but is designed a lot more holistically than VSCode.

Zed to me feels like a great batteries-included editor and I still run it as my non-emacs alternate editor. I wish its configuration was a bit more discoverable (especially with configuring linters/formatters), but it's 95% of what I need 95% of the time.

I had to use VSCode for some projects in the past because it was what was available on the clients workstations... I can't imagine having to use that laggy electron abomination all the time. For me Zed is sent from heaven, because my previously preferred editor (geany) hast basically zero developtment nowadays.

Zed succeeds at reducing the switching cost. I used NeoVim for ten years daily and configured it way back in college days.

I thought I would be unable to move to a GUI editor and it turns out that the speed and efficiency of Zed plus the almost one-to-one mapping of Vim features means that I am extremely productive in Zed.

I think the point of ACP being an open protocol is so that other editors (e.g. VSCode, Neovim) can implement it as a receiver and integration with ClaudeCode/GeminiCLI/... would just work.

OTOH I'd ten to prefer as less plugins as possible in VSCode, just because they are inherently dangerous, I'd like Zed plugins that are WASM, so they don't have access to the world.

But I agree that VSCode Typescript support is better than Zed, it works with weird projects setup, while Zed has more troubles. I at work VSCode and Zed/Helix for my projects, generally I use Zed when want to do some AI stuff, otherwise I just use Helix.

It sounds like you haven't tried Jetbrains IDEs. I understand still preferring VsCode in that case, but I think you would be saying "I prefer VsCode" vs "I don't see a reason". A big con with JB is they are very slow. The upside is that they manage multi-file projects, refactoring, and introspection far better than VsCode.

I switched to Neovim a year ago, and while I did spend a significant amount of time on configuring it, I haven't touched my config for months now - and I'm perfectly happy with it. There's things I can improve, but it does what I want.

I am sure some would have said the same about why would anyone switch to using Linux when there was Microsoft Windows.

I use zed when I need to quickly edit something, it pops up so fast- everything else feels sluggish.

Plausible take. My first time using a coding agent was with Zed’s. As a sceptic, I was blown away. Two or weeks later Claude Code dropped and I haven’t used the Zed agent since. I canceled my subscription because 20 EUR/month for AI features I no longer use felt silly. And with the massive investment from Sequoia, and the real possibility of enshittification, I couldn’t mentally classify it as a monthly ‘donation’ to sponsor ongoing development either.

That last part is just me maybe. But I’m sure they poured a lot of time into their (granted, very nice) agent, only for many users to switch to Claude shortly after release. Letting others build the agent but being the place where the agent get used seems like a smart move

I think the article mentions this not yet being supported by the SDK, and therefore by the editor.

The post has a video included. More integrated interface seems useful.

You get a diff view of the edits CC is making to your project and there's a mode where you editor's focus follows where CC is editing in your project.

If you're going to live in the terminal, please do yourself a favor and replace Terminal.app with iTerm2, and replace bash with zsh.

Actually, Anthropic built that integration themselves.

Does it? Or do you mean the integration when you run Claude Code from the terminal opened in VSCode? (which is already pretty good)

I wish it natively supported vimrc, I dislike having to rewrite all my bindings in zed format.

After having troubles with Claude desktop not properly updating canvas document, I found that apparently Claude Code is same as desktop one plus you can use it not just for code but documents in VS Code.

In Elixir/Rails (and newly React) world you can connect Clade Desktop to Tidewave, which gives you pretty close integration with codebase and CLI tools. That makes the experience somewhat similar to running Claude Code in editor's terminal, or now directly connected in Zed.

I think basically all these tools are trying to integrate all the stuff together and find the best way to do that, so that you don't need to be copy-pasting stuff around, jumping between tools, etc.

Not tried with Zed yet, but in VS Code it's mostly the same except you can view the diffs in the editor, and can give context from the open file. I think it can also see errors from the editor.

Not just your codebase, but also a wide range of *nix tooling. Night and day to working on the CLI vs an IDE.

They've been chasing nothing but AI crap for months. Maybe a year. They've stopped improving the core editor completely.

I'm getting ready to just switch back to Sublime Text

There is a Windows version if you're willing to use unofficial builds. I've been using it successfully for a few weeks. There's a few more rough corners than in the official Mac/Linux version, but it's usable.

Am I reading this wrong, they added support for Claude Code, which should work with your Anthropic sub. The Claude via API that has always been there is not what they're talking about in this anoucement

We just pushed a fix! Here's how to get it:

- Start a new Claude Code Thread, which will kick off the background download of the new Claude Code ACP adapter.

- Wait a few seconds, then start another Claude Code Thread. The new thread will use the updated one.

We're working on a nicer UX for getting updated versions, which we'll definitely ship before Claude Code support leaves beta!

I got this same error. I restarted Zed, and it fixed itself.

I got this error too.

Tried restarting zed and restarting my computer, but neither resolved it.

Same!

Is it really comparable?? I use JB IDEs heavily and when I tried Zed a few months ago it felt more like Notepad than IntelliJ. Maybe that’s what you’re looking for and I am sure that’s fast, but they seem to have quite different use cases.

After moving to Zed I also missed good Git integration that I had in VSCode with GitLens.

That pushed me to finally try Jujutsu instead of Git, since it has a CLI (and TUI and GUI, if that's your thing) that are perfectly usable by mere mortals. Now I feel as fast as when I had the Git integration, despite using the terminal.

But if I had stuck to Git, trying GitButler or Sublime Merge was the next option on my TODO list.

I've been using with a subscription, both API and subs work.

When you run Claude Code it lets you pick between API key and Claude subscription login. It sounds like it found an API key in your env and is trying to use that. You may have to find a way to trigger a login to get you into your subscription account.

I actually works already, just add new agent definition:

--- "Local QWEN": { "command": "/usr/local/bin/qwen", "args": [ "--experimental-acp" ], "env": {} }, ---

Zed "just works" with ollama, so if you install a qwen modal locally you're set.

Is that much different from just using the Zed agent with Qwen via API?

agreed. waiting for qwen here as well.

qwen keep rebase gemeni so I guess just wait?

It will come! Linux support is only recently getting good. They'll get there.

The editor treats edits from Claude Code as a first class citizen. You can easily review, approve, rollback, etc. Claude's changes in a curated experience that is much faster than digging around in diffs or needing to approve each edit as it is proposed.

https://zed.dev/agentic

It's just about polish and tightness of integration or may even be lack of it. Claude Code for VSCode is basically Claude Code running in VSCode terminal with some integrations for open file, selections and diffs.

This is why I try to avoid building/supporting desktop Linux gui apps

I know what you mean. I used to like Emacs but then they added syntax highlighting for Rust and I don't use Rust so I'll never touch that editor ever again.

Just disable it. It seems like they've made is as simple as possible to do so while also being realistic about what it takes to make a successful (read: profitable) IDE the year 2025.

It was always going to be infested with something or other, you don't raise $42M for just a text editor. Could be worse, a few years earlier it probably would have incorporated blockchains somehow.

How does the existence of LLM features impact the quality of the rest of the editor? It's a single toggle to disable everything [1].

[1] https://zed.dev/blog/disable-ai-features

One line fix for you…

https://zed.dev/blog/disable-ai-features

This seems obviously untrue. https://zed.dev/extensions

I recently un-borked my emacs config and re-downloaded elisp into my brain so I'd be happy to help walking you through the emacs stuff. I haven't used emacs on Windows natively in a long time but I did it for years and found it worked quite well so you may want to try that again.

FWIW emacs these days has native LSP support (eglot) and native tree-sitter supported, but you'll need to grab tree-sitter libraries and LSP servers. For MacOS I found just using brew to install most of these works great, and I wonder if Windows can't work the same way. If you're in a VM you should just be able to have your package manager do the work.

LLM support in emacs is still a bit primitive but there are packages like emacs claude-code that are changing things. Personally I wrote an elisp function which grabs the filename of the emacs window relative to the project root and copies it so that I can just paste it into claude code and have it ask questions or do things.

```

(defun copy-buffer-file-name-from-project-root () "Copy the relative path of the buffer in this project to the kill ring" (interactive) (kill-new (file-relative-name (buffer-file-name (current-buffer)) (project-root (project-current)))))

```

if you're curious but depending on how you have your clipboards setup with the kill ring, you may need to modify this.

---

As far as fighting the JS fight on the Zed website, well, I think this thread isn't one of those upvote-anti-JS-rant threads that spring up on this site all the time so shrug.

Lucky you because for me, for some reason, their website is randomly down when I click to change page: Web server (host) is returning an unknown error Error code 520.