At Olostep, we are building Web Infrastructure for AI.
The primary user of the web is switching from humans to AI agents. Today, billions of people use the web. Tomorrow, trillions of agents will.
Olostep was born on the bet that eventually all knowledge work will flow through models and that there will be a myriad of AI application companies sprouting in different industries to make this happen.
This is what's started to happen with the pioneers of this era. We now have AI for Marketers: AirOps, Profound, Petralabs. AI for engineers: Cursor, Devin, Factory. AI for Lawyers: Legora, Lexroom, Harvey. These companies are deploying AI agents that need the web to assist and automate knowledge work.
Why do they need the web?
Over the past 30 years, the web has become the collective source of humanity's knowledge and where work happens. It has become the OS on which the world runs. Without access to this living corpus, AI systems can only experience the past.
Olostep is building the underlying agentic search infrastructure they all need.
The portal to the Web for humans was a search engine. For AIs, it will be an API.
At Olostep (www.olostep.com), we started with a suite of Web Data APIs for AI agents to search, extract, and structure Web Data.
Olostep APIs are used by agents deployed in some of the fastest-growing companies in the world. These companies are ushering humanity in a new era of abundance and progress. Olostep exists to make the world's data useful to the businesses pushing humanity forward
The team and our conviction
Arslan (CS@Stanford, fmr lead engineer at 2 YC startups) worked on this problem in the early days at Legora, the Swedish legal-tech company now valued at $5.6B, and then together with Hamza at Zecento, one of the most popular AI-powered e-commerce products in Italy.
We're at an inflection point similar to the late 90s internet, but the scale is different. If the first web connected billions of people to information, the next connects trillions of agents to knowledge. Each agent needs to search, extract, and structure data at volumes humans never could.
We're building the infrastructure that makes the web accessible to AI systems to automate the real-world economy. Our conviction is straightforward: agents need access to real-time web data to be useful, and that access needs to be reliable, structured, built for how models actually work, and contextualized for each vertical. Indexing the whole web for agents is not the right path forward. Agents need specialized search engine with memory to run business queries.
What's Next
When the primary user of the web changes from humans to agents, this disrupts the economic incentive on which the internet was built. Today's web runs on attention and ads. Tomorrow's web will run on access and utility.
Companies like Google make around $300B from ads alone. And now the company building the agentic API gateway to the world's data will get to decide and benefits from this monumental shift.
Another inextricably linked area to the industry in which Olostep is operating is: What will be the interface for this? For humans, browsers have become synonymous with search and web access. We use the browser both to search and extract data from the Web but also to work and act on it (e.g. apply for jobs, submit forms, buy things online). Will the same apply to AIs?
Will AIs use APIs or a browser? or both?
As we transition from a Web that simply answers our questions to an agentic Web that anticipates our needs, entirely new possibilities and scenarios emerge.
We believe search and data extraction for AI will be an API first approach, like the one we have been focusing on at Olostep with our unique architecture and insights.
Instead, for working and acting on the web, the browser for AI agents will look different from current solutions, especially for business applications. Today's options—headless browsers in the cloud and consumer browsers like Comet or Atlas—miss key requirements.
Thinking from first principles, for AI to do work on our behalf, especially in a business context, it needs to:
- Operate where the user is: On the edge to leverage hardware advancements (e.g. M5 chip), guarantee privacy and reliability, be cost effective, have low latency and provide access to user cookies and context (something headless browsers in the cloud cannot provide)
- Preserve user freedom: By working within users' existing browsers rather than forcing them to switch environments or lock into single-vendor ecosystems (the approach taken by Comet or Atlas, which can work for consumers but not businesses)
- Be extensible: Every business has unique edge cases to handle. They need foundational building blocks that they can customize for their specific requirements—not rigid, one-size-fits-all solutions
We have been working closely with startups and companies on this, and it's clear that Olostep Orbit is the right solution. We are going to launch it in 2027.
Our goal for the next decade is to build the web operating system for global business:
Data-as-Infrastructure → Agentic Search → Web OS for Business