We build the agent layer for business websites
Sitora exists because the answers customers want already exist. Businesses wrote them, carefully, on their own sites. Someone just had to make those answers able to speak.
Sitora started in 2024 in Denver, after Maya spent a year watching session recordings for SMB software customers. The same scene repeated thousands of times: a visitor lands, opens two pages, searches for one specific thing, and leaves. The answer was on the site. It just was not going to introduce itself.
Modern language models changed what was possible. Reading a whole, messy, human-written website and speaking for it faithfully, with sources, in the business's own voice, is not a chatbot script problem. It only became solvable in the last few years, which is why we work on it full time, and only on it.
We are a small US team that believes the winning agent will be the most honest one. Every design decision at Sitora starts from the same rule: a wrong answer costs a business more than no answer. That is why grounding, visible sources, and the handoff are the product, not features of it.
Two people who kept meeting the same problem
Maya Sandoval
Maya spent seven years running growth and support tooling for SMB software companies, most recently leading the self-serve business at a Denver point-of-sale platform. She started Sitora after watching thousands of session recordings where visitors searched a site, found nothing, and left.
Daniel Brook
Daniel built retrieval and ranking systems for a decade, including search infrastructure used by millions of storefronts. At Sitora he is responsible for the grounding pipeline: how an agent reads a site, decides what it knows, and proves where every answer came from.
Around them: seven more across engineering, design, and support, all US-based, all full time on Sitora. We are deliberately small; agents should scale, headcount should not have to.
We hire slowly and deliberately
Three roles are open right now.