N
Show HN
Show HN: Build Web Automations via Demonstration
Hey HN,

We’ve been building browser agents for a while. In production, we kept converging on the same pattern: deterministic scripts for the happy path, agents only for edge cases. So we built Demonstrate Mode.

The idea is simple: You perform your workflow once in a remote browser. Notte records the interactions and generates deterministic automation code.

How it works: - Record clicks, inputs, navigations in a cloud browser - Compile them into deterministic code (no LLM at runtime) - Run and deploy on managed browser infrastructure

Closest analog is Playwright codegen but: - Infrastructure is handled (remote browsers, proxies, auth state) - Code runs in a deployable runtime with logs, retries, and optional agent fallback

Agents are great for prototyping and dynamic steps, but for production we usually want versioned code and predictable cost/behavior. Happy to dive into implementation details in the comments.

Demo: https://www.loom.com/share/f83cb83ecd5e48188dd9741724cde49a

-- Andrea & Lucas, Notte Founders

We're full circle to "Programming by Demonstration" (1993) https://acypher.com/wwid/ or Pygmalion (1975), though this iteration probably works better :)
Full circle indeed! We hope this iteration works better :)
I used IBM CoScripter[1] for many many years to supercharge productivity at work. It records macros and you can rerun it. Perfect for most Corporate applications which require you to fill many pages of garbage to close tickets/etc. I used to close hundreds of tickets/day working as a Linux production support engineer. The troubleshooting was quick, closing the tickets on the bullshit peoplesoft ticketing system with a dozen screens was the most timeconsuming thing. CoScripter helped me quite a bit. I've searched high and low for simple macro recorders like that, but never found any. I wonder if this solves the same problem.

https://en.wikipedia.org/wiki/CoScripter

Same spirit, yes. Record once, replay many. The difference now is when the page changes or an element moves, we have those fallback AI agents that can recover instead of just failing. But the core idea is the same; automate browser stuff.
Why is this not a Launch YC (or at least mention it?) since you seem to be part of the current batch?

The record/replay is definitely and interesting direction. The browser automation space is getting super crowded though (even within YC), so curious to hear how you differentiate from:

- BrowserUse

- Browserbase

- BrowserBook

- Skyvern

We're YC S25, launched in summer. Demonstrate Mode is a new feature we recently added to our platform and thought it would be worth sharing here.

Re differentiation: The space is crowded and feature sets converge. But like LLM providers, we feel there's room for multiple players with different positioning long term (enterprise, developers, etc.) Right now, we're now focused on making the product that feels most exciting to build with - hope people can tell that :)

I do this with Chrome recording and Playwright. What I need is an AI agent to meander through my product as if it were the target user and test/break things so I can pass that to my LLM to fix. Does anyone have that?
Different from our core use case, but our agents can do open-ended exploration as well. You could prompt something like "navigate to this app as a new user and try common. flows" with structured outputs for findings. Session recording will show what happened. Not sure if it fully solves your problem - but happy to explore this together if you want to try it.
[dead]
Using this, if a first time user logs in, could we share automated scripts, that they can execute to create sample workflows?
Yes. Record an automation flow, export the code, and share it. New users can run it as-is on our infra, or pick modify and run elsewhere.
Congrats on the launch and nice identity. You might want to check Selenium as a source of inspiration
Thanks :) We're Playwright / CDP under the hood
pardon me but arent there like 5000 startups that do this exact same thing
There are a few yes
Interesting but are there any open source products which can do the same too? This does feel something that I can trust more if it was open source personally but good luck with the product!
The core Notte framework is open source (SSPL): https://github.com/nottelabs/notte. Only the managed infra is hosted. You can also use our Demonstrate Mode as a way to bootstrap a script and then run it on your machine if you prefer :)
Browser use has a project "Workflow Use" that has similar aims: https://github.com/browser-use/workflow-use
Similar goals but quite different execution and primitives
Playwright codegen.
Codegen hosted in a way. Codegen is local and outputs a Playwright script. We add managed infra, observability, retries, and agent fallback recoveries when things break.