UU API is built by OpenDataLabs.
OpenDataLabs is an applied research institute. We develop technology that helps AI understand the humans using it. UU API is the Context Gateway, renamed, redesigned and opened up as a developer platform.
AI gets better when it understands the human in front of it. The way to get there is to ask that human, once, clearly, and keep the receipt.
Built on an open standard
UU API speaks PDPP, the Personal Data Portability Protocol: an authorization and disclosure protocol for personal data. It is developed in the open at PDP-Connect, a lab at Linux Foundation Decentralized Trust, as a neutral, vendor-independent home for the specification and its reference implementation.
Every connector is a signed, versioned artifact in a public registry. Anyone can read the schemas, run the tests, or open a pull request.
- Owner authorizes
The user authorizes a client to read a bounded, purpose-scoped slice of their data, for a stated purpose and duration.
- Source is collected
A native provider or a polyfill connector collects the source into a single queryable substrate, the user's Personal Server.
- Client reads, with a receipt
Every read is bound to the grant. Revoke the grant and reads stop. The consent chain is auditable end to end.
Case studies
All 10 case studiesThweet creates 1,497 one-of-one guest experiences.
Spotify listening history became a custom whiskey ice cream flavour and a plushie for every guest at Korea Blockchain Week 2025.
Frontier AI Lab sources real human conversation data with consent, redaction and provenance.
Thousands of contributors, one permission event each, two redaction passes, a consent record on every row.
Avinasi Labs accelerates longevity research through direct data access.
Health data from consenting contributors, direct to the research pipeline.
Solo AI fine-tunes a generative music model on real listening intent.
Ground-truth listening data instead of scraped playlists.
Mavu solves the matching problem with better user context.
Micro-workers matched on what they actually do, not what a form says.
Fortune 500 fashion brand builds a real-time taste prediction engine.
Predicting style preference without purchase history, from tens of thousands of consenting contributors.
Example apps, forkable
Sleep Coach (Oura), Habit Mirror and Spending Nudge (iCloud Notes), Personal Shopper and Taste Profile (Instagram). All on GitHub under vana-com.