See how anything was built, then have your agent build it. Powered by @reatools Ebdow8824bQWcVDTGh1nyitfGrBrtTfARRDpq5VYhSwL
Joined October 2026
- Tweets25
- Following10
- Followers83
- Likes2
Pinned Tweet
Introducing Engineer Anything.
Enter any URL and get a full technical breakdown of how the site was built: DOM structure, JavaScript bundles, network activity, API routes and storage, returned as a single markdown file.
Free to use, open source, and available as an API for agents.
Powered by @reatools
enganything.com
github.com/engineer-anything…
By doing this, Engineer Anything becomes the only place to build a website for free, and actually well.
Show it the sites you like. Because it's built on @reatools, it understands exactly how they work, down to the detail, and generates an almost perfect site based on what you want.
Then our agent, also powered by @reatools, builds it for you. Free, funded by creator fees.
Here's where Engineer Anything is heading.
Right now, a scan hands you the full structure of any site: the DOM, the scripts and their dependencies, the API routes, the network calls and the storage. It's everything about how the site is put together, returned as one markdown file. That's genuinely useful, but there's still a step left for you: reading through it and working out how to actually rebuild it.
That's the step we want to remove.
The goal is for scans to come back not as a breakdown to interpret, but as an entire build. Instead of raw data, you'd get a clear plan: the framework the site runs on, detected automatically. How the pieces fit together, from the layout down to the components. Which API routes power which features. And the order to build it all in, so nothing's left for guesswork.
Hand that to your agent and it has everything it needs to go from a single URL to a working clone in one shot. No decoding, no filling in the gaps, just a plan it can run with start to finish.
This is where creator fees come in. They would allow us to fund the AI work behind turning a scan into a full build plan, while keeping Engineer Anything free and open source for everyone.
Here's where Engineer Anything is heading.
Right now, a scan hands you the full structure of any site: the DOM, the scripts and their dependencies, the API routes, the network calls and the storage. It's everything about how the site is put together, returned as one markdown file. That's genuinely useful, but there's still a step left for you: reading through it and working out how to actually rebuild it.
That's the step we want to remove.
The goal is for scans to come back not as a breakdown to interpret, but as an entire build. Instead of raw data, you'd get a clear plan: the framework the site runs on, detected automatically. How the pieces fit together, from the layout down to the components. Which API routes power which features. And the order to build it all in, so nothing's left for guesswork.
Hand that to your agent and it has everything it needs to go from a single URL to a working clone in one shot. No decoding, no filling in the gaps, just a plan it can run with start to finish.
This is where creator fees come in. They would allow us to fund the AI work behind turning a scan into a full build plan, while keeping Engineer Anything free and open source for everyone.
Engineer Anything just passed 300 scans.
300 URLs turned into full reconstruction plans: DOM, scripts, API routes, network calls and storage, all handed back as markdown ready for an agent to build from.
Thanks to everyone who's tried it so far. This is just the start, and it stays free and open source.
Engineer Anything update.
Since launch, the tool's been doing exactly what it set out to do: take any URL and hand back a full reconstruction plan your agent can build from. DOM, scripts, API routes, network calls, storage, all in one file, in under a minute.
And it's only the start. With creator fees going back into the project, here's what we're working towards:
- Full-site scanning across multiple pages instead of one
- AI-written build plans that turn raw scans into clear instructions
- Automatic framework detection
- An MCP server so Claude and Cursor can use it directly
- Higher API limits and faster scans
It stays free and open source throughout. More soon.
Following on from the $400 in fees, here's what each upgrade would actually bring to Engineer Anything:
- Dedicated scanning server: scans run faster and don't queue up when a lot of people are using it at once, so you get your results quicker
- Higher API limits: the extra capacity would allow us to raise the cap, so your agent can run far more scans before it has to wait
- AI model credits: lets us start testing written build plans, turning the raw scan into a clear "here's how this site works and how to build it" summary your agent can follow more easily
- Uptime monitoring: we'd know the moment the API goes down and can fix it fast, so it's there when you need it
- Scan history storage: your past scans get saved so you can revisit them and share results by link instead of re-running them
Quick update: so far the token has generated $400 in creator fees, and all of it is going back into Engineer Anything.
Here's what we intend to put it towards:
- A dedicated server for scanning, so scans run faster and don't queue up when traffic picks up
- Higher API rate limits, which the extra capacity would allow us to raise
- AI model credits to start testing written build plans generated from each scan
- Uptime monitoring so we know straight away if the API goes down
- Storage for saving scan history so results can be shared by link
As fees come in, we'll keep posting updates on where the money goes. Nothing is final and plans could change based on what people use most, but the goal is to make Engineer Anything faster and more useful while keeping it free.
enganything.com/
Quick update: so far the token has generated $400 in creator fees, and all of it is going back into Engineer Anything.
Here's what we intend to put it towards:
- A dedicated server for scanning, so scans run faster and don't queue up when traffic picks up
- Higher API rate limits, which the extra capacity would allow us to raise
- AI model credits to start testing written build plans generated from each scan
- Uptime monitoring so we know straight away if the API goes down
- Storage for saving scan history so results can be shared by link
As fees come in, we'll keep posting updates on where the money goes. Nothing is final and plans could change based on what people use most, but the goal is to make Engineer Anything faster and more useful while keeping it free.
enganything.com/
Engineer Anything is an open source tool that shows you how any website was built. You give it a URL, and it returns a full reconstruction plan as a markdown file that you can read yourself or hand to your agent to build from.
Each scan covers:
- DOM structure and tag breakdown
- JavaScript bundles, file sizes and dependencies
- Accessibility tree and page content
- Network requests and API routes
- localStorage, sessionStorage, IndexedDB and caches
- Screenshots of the layout
You can scan the full site or focus on just the API, the UI or the JS bundles, at quick, normal or deep depth. All analysis is passive, so nothing on the target site is executed or modified.
There are three ways to use it: in the browser with no setup, locally from your terminal, or through our free API so your agent can scan sites and start building on its own. It's built on @reatools and MIT licensed.
The project is funded through @pumpfun creator fees, which go back into development. That would allow us to keep it free while working towards faster scanning infrastructure, higher API limits, full-site crawling, AI-written build plans and an MCP server for Claude and Cursor.
Site: enganything.com/
Docs: github.com/engineer-anything…
Engineer Anything also records what a site stores in your browser: localStorage, sessionStorage, IndexedDB and caches. It's often the quickest way to see how an app handles state and sessions.
Each scan includes the accessibility tree, which shows how the page is structured semantically and what text it contains. It's useful for agents, since it gives them the meaning of the page rather than just the raw HTML.
Engineer Anything only observes. It loads the page, reads what's there and records what happens, but nothing on the target site is executed or modified. It's for understanding how things are built, not breaking them.
Every scan has three depth options. Quick gives you a fast surface pass, normal watches the page for a few seconds per phase, and deep spends longer observing so it picks up more network activity and data.
You don't always need to scan everything. Engineer Anything lets you choose a full scan, just the API surface, just the UI and design, or only the JavaScript bundles, depending on what you're trying to learn.
engineer Anything has a free API. Your agent sends a URL, checks back for the result, and gets a full reconstruction plan as markdown. No API key needed. Examples for Python and JavaScript are in the docs.
If you like how a landing page is put together, you can scan it with Engineer Anything, give the markdown file to Claude or Cursor, and ask it to build you something similar. It's a much faster starting point than describing it from scratch.
The way people build software is changing. More and more of it is being done by agents, and agents work best when they have a clear picture of what they're building. That's exactly what Engineer Anything gives them: a full breakdown of how any website was built, from a single URL.
It's early, but the foundation is there. Creator fees would allow us to invest in:
- Dedicated scanning infrastructure for faster scans with no queues
- Much higher API limits for agents
- Full-site crawling instead of single pages
-AI model credits to turn every scan into a written build plan
- Automatic framework detection
- An MCP server so Claude and Cursor can use it as a built-in tool
- Saved scan history you can share by link
Engineer Anything lets you scan any website and get back a full breakdown of how it was built: the page structure, scripts, dependencies, API routes, network calls and storage. It all comes out as one markdown file that your agent can read and start building from straight away.
You can use it in the browser with no setup, run it locally from the terminal, or connect your agent directly to our free API so it can scan sites by itself. It's built on @reatools and fully open source.
Want to run Engineer Anything from your own terminal or agent?
Clone it and run it locally in three commands, or use the free API to connect it directly to your agent.
Everything's in the docs:
enganything.com/docs
How Engineer Anything works:
1. You enter a URL and choose a scope (full site, API, UI or JS bundles) and a scan depth.
2. A fresh Chrome instance opens the page and waits for it to fully load.
3. Using @reatools, it inspects the DOM and accessibility tree to map the page structure and content.
4. It analyses the JavaScript bundles to identify scripts, file sizes and dependencies.
5. It observes network activity to detect API routes and requests, and records localStorage, sessionStorage, IndexedDB and caches.
6. It captures screenshots of the layout.
7. Everything is compiled into a structured markdown reconstruction plan, ready to hand to your agent.
All analysis is passive. Nothing on the target site is executed or modified.
Use it in the browser, run it locally, or connect your agent directly through the API.
Engineer Anything will be funded entirely through @pumpfun creator fees, with every fee reinvested directly into the project.
These fees would allow us to keep Engineer Anything free for everyone, while funding upgrades that could make it a far more powerful tool:
- Expanded scanning infrastructure for faster results
- Higher API limits so agents can run more scans
- Deep scans as the default
- Full-site scanning across multiple pages
- AI-written build plans from every scan
- An MCP server for Claude, Cursor and other agents
- Saved and shareable scan history
Free, open source, and built to grow with its users.
Powered by @reatools
enganything.com
github.com/engineer-anything
Ebdow8824bQWcVDTGh1nyitfGrBrtTfARRDpq5VYhSwL
