Show HN: Stagewise – frontend coding agent for real codebases | Hacker News Reader

Show HN: Stagewise – frontend coding agent for real codebases

stagewise.io

42 pointsglenntws18 comments

Hey HN, we're Glenn and Julian, and we're building stagewise (https://stagewise.io), a frontend coding agent that inside your app’s dev mode and that makes changes in your local codebase.

We’re compatible with any framework and any component library. Think of it like a v0 of Lovable that works locally and with any existing codebase.

You can spawn the agent into locally running web apps in dev mode with `npx stagewise` from the project root. The agent lets you then click on HTML Elements in your app, enter prompts like 'increase the height here' and will implement the changes in your source code.

Before stagewise, we were building a vertical SaaS for logistics from scratch and loved using prototyping tools like v0 or lovable to get to the first version. But when switching from v0/ lovable to Cursor for local development, we felt like the frontend magic was gone. So, we decided to build stagewise to bring that same magic to local development.

The first version of stagewise just forwarded a prompt with browser context to existing IDEs and agents (Cursor, Cline, ..) and went viral on X after we open sourced it. However, the APIs of existing coding agents were very limiting, so we figured that building our own agent would unlock the full potential of stagewise.

Since our last Show HN (https://news.ycombinator.com/item?id=44798553), we launched a few very important features and changes: You now have a proprietary chat history with the agent, an undo button to revert changes, and we increased the amount of free credits AND reduced the pricing by 50%. We made a video about all these changes, showing you how stagewise works: https://x.com/goetzejulian/status/1959835222712955140/video/....

So far, we've seen great adoption from non-technical users who wanted to continue building their lovable prototype locally. We personally use the agent almost daily to make changes to our landing page and to build the UI of new features on our console (https://console.stagewise.io).

If you have an app running in dev mode, simply `cd` into the app directory and run `npx stagewise` - the agent should appear, ready to play with.

We're very excited to hear your feedback!

Open articleView on HN

Pet peeve because everyone seems to be doing this:

The video demo shows the operator asking for "same height", and then the agent sets a hardcoded 298px height instead of using any of the many reliable/solid ways to do it with CSS.

I'm sure there are better examples you could prompt to make that video convincing, though I appreciate the honesty showing what to expect.

Good catch - we could definitely put more effort into creating demos

feels like a case of a system prompt should be enforcing the AI to avoid usage of arbitrary numbers like specific pixel counts and to leverage modern css standard for sizes.

that gave the best results so far - tweaking the system prompt to make sure the agent respects the project's ui system, is aware of dark mode and responsiveness, etc.

Before I saw this comment I thought that's how it "fixed" it. I recorded my screen so I could slow it down and see what it said and sure enough:

    I've made the following change:
    1. Added a fixed height of `h-[298px]` to match the tallest card (which was 298px)

Screenshot: https://cs.joshstrange.com/d75pC236

This joins the ranks of prominent demos on the landing page that do the wrong or silly thing. It really makes it hard to take some things seriously.

That said, I like the overall idea of this. I think that Claude Code and friends will get way more powerful as we find ways to feed them better (not necessarily more) context.

More context is not going solve this.

CSS requires understanding emerging properties of layout. So context doesn't help. Knowing how a variable expands a a div into a set of constraints isnt context dependent.

What you probably mean is a well designed model context which fits the most flexible and logical code framework.

Reading the last paragraph should make you realize the proposition isnt likely going to happen.

Exactly, but do u think a model can be developed to deal with this?

Unrelated to the actual product (which looks nice), I'm always curious when I see these landing pages with logos from massive companies on them. I'd bet a beer that you didn't go through Oracle's processes for using their logo, for example. Are you just YOLOing it and hoping they don't come after you for unauthorised use?

tbh, we just looked at our users/ stargazers and picked the most interesting companies - and were cautious with the wording ("embraced by engineers")

so, i guess YOLOing it is the best way to put it

In other words trying to be as deceptive as possible while being able to CYA

Came here to sat same. Airbnb has strict policies around what they endorse. Even things used can't be talked about. Just as FYI it doesn't come off great to folks that make decisions about future usage to see this

great point, we haven't really spent too much thoughts about that.. changing the logos to companies we know well makes more sense, we'll do that to save us trouble

Cool product! I have a question outside of this topic: how did you manage your open-source processes from day one, and how did you encourage people to contribute to the project?

We started with a setup that was quite GitHub-friendly (we set up https://github.com/changesets/changesets and added a small contribution guide).

People then found the courage to contribute themselves. We haven't really pushed for contribution so far!

I really liked the look of the product, but since the founder replied multiple times to questions / comments some thing along the lines of “we have not really spent time on…” comes across not very well thought out!

However I am personally looking for a product exactly like this! I really want LLMs to make beautiful UI for me.

Weird how you can’t sign up except via the CLI. And pricing isnt posted publicly and usage isn’t documented. The docs leave something to be desired as well. Would be nice to show more info for running this with a backend process as well

I can't find info on your GitHub pages how dataflow works in this tool.

You can show an example of what is looking for and what is actually being sent to LLM

this should be sandboxed by default, running agents directly on the host is a horrible idea.

Reply on news.ycombinator.com