Companies want AI on their data. Their data is not ready.
Every company wants its people to use AI safely on what the company knows. But that knowledge is messy and scattered across the tools where work happens, and as it sits, nobody gets value from it. Mithrandir delivers it secure and organized, inside every AI agent people already use, with each person seeing only what they can access at work.
- Founded
- 2026
- Research
- CMU · LTI
What we believe
Company data is worth little as it sits. Scattered across chats, mail, documents, tickets and calls, duplicated and out of date, it cannot answer a question, and most of it is never found again.
Using it with AI has to be safe. Each person should see only what they can already access at work, with the company's hierarchy deciding where a tool has no rules, and the data should stay in the company's own cloud.
Nobody should switch tools. The memory belongs inside the AI agents people already use, organized into one structure, so every team gets value from what the company already knows.
Every company already knows more than it can use. We make it usable, safely, for everyone.
Built alongside ongoing research.
Our work is informed by Salih Mert Uzun's ongoing research at Carnegie Mellon University's School of Computer Science, Language Technologies Institute, on the limits of determinism and rule-following in large language models.
Who's building this
Emir Akar
Co-founder
Built Mithrandir's entire web platform, the site you're looking at right now. Previously built enterprise systems across multiple projects.
LinkedIn →
Salih Mert Uzun
Co-founder
Leads Mithrandir's system architecture and design. Built and sold a patent-pending product to 10+ venues across the US.
LinkedIn →
See it on your own work.
Bring one workflow where company knowledge keeps getting lost or stays out of reach. Thirty minutes with the founders, no form to fill in first.
Book a call