Infrastructure
Automated Web Reader
The site changed its layout. Your data should not vanish.
A tool that reads pages on the open web for an AI assistant, gets past the blocks that stop most of them, and finds the information again after the website is redesigned.
Where it runsOn a rented server, so any AI agent can call it from anywhere. One person can also run it on their own machine.
Who is this for?
People running AI agents that need to read the web. Before this, one redesign somewhere else quietly emptied their results.
What the numbers mean
Cost metering and self repair. Everything else already existed and was not rebuilt.
Including one that runs the whole tool end to end, the way an AI agent would actually call it.
How it actually works
Anything that reads information off a website is aimed at a specific spot on the page. The day that site is redesigned the spot moves, nothing reports an error, and the results quietly come back empty.
Empty is the dangerous answer, because it looks exactly like a working system that found nothing. This goes and finds the field again in the new layout, and tells you where it moved to.
It also gets past the checks that block automated readers, and it keeps a copy of pages it read recently, so asking the same question twice only costs once.
Every request is counted and priced as it happens, so you can see which agent spent what, and how much the stored copies saved you.
It is called a prototype here because that is what it is. The meter counts money accurately and there is still nothing attached to it that bills anybody.
A data feed that breaks loudly costs you an afternoon. One that breaks quietly can run wrong for months.
We build things like this. Tell us what you are stuck on.
What it is built with
- Python
- MCP
- FastAPI
- Playwright
- Docker