Revenue share
You already hold the rights. We license accepted data and share the proceeds.
AuraOne / Data Partnerships / Agent traces
Your agents do real work. That work can become training data.
We help AI builders find out what their traces are worth, prove what they can license, and bring them to the teams who need them.
Opt-in only. You choose what is included. No files, code, or production access are needed to start.
- The task“Customers are being charged twice. Find it. Fix it.”
- It investigatesReads the payment code. Checks the logs. Spots a duplicate transaction.
- It tries a fixChanges the code. Runs the tests. They fail.Test failed
- It corrects itselfReads the error. Fixes the fix. Tests pass.Recovered
- The outcomeA developer approves it. It ships. The double charges stop.Verified
The value is not the final answer. It is the failure, the recovery, and the proof.
What counts as a trace
Not every trace is equal.
Depth and proof decide value. A short, verified trace can be worth more than a long conversation nobody checked.
- ConversationA question and an answer. Useful, but it does not show the answer was right.
- ExecutionEvery tool call, failure, and retry. How the work was attempted.
- VerifiedThe steps plus proof: a passing test, an approved edit, a confirmed result.
- ReplayableThe task rebuilt as an environment another model can attempt again and again.
What you might already record
- Prompts and tool calls
- Code edits and test runs
- Browser actions
- Failed attempts and retries
- Approvals, rejections, and accepted results
A gateway may see prompts and responses but not whether the code worked. That is still worth a conversation. Outcome evidence is what moves a trace up the list.
Who this is for
Built for AI apps, coding tools, and gateways.
AI apps. Coding tools. Inference gateways. Teams whose engineers work with AI agents. A few months old is fine. Volume and outcomes matter more than age.
How partners earn
Three ways to earn from the work.
Managed contribution
We help you design an opt-in program, set quality gates, and find buyers.
Buyer-funded collection
A model builder pays for a specific dataset or environment, scoped up front.
Value is counted in accepted traces, not raw logs. Every term is set in a signed agreement. Payment is never guaranteed.
Privacy and ownership
Your users’ trust comes first.
We start with a conversation about what you record, which rights you hold, and which outcomes you can show. Nothing moves until you approve exactly what moves.
Opt-in only
Only sessions your users, customers, or team have agreed to share are considered.
You set the boundary
Customer content, employee content, secrets, and anything you exclude stays out.
Rights come first
Permission to train your own model is not always permission to license. We check before anything else.
From a first conversation to a licensed dataset.
Each step has a decision owner, a rights boundary, and a written result.
01
A short conversation
What your product does, what it records, who can authorize licensing, and what happens after the agent finishes.
Business context only
02
Rights and sample review
We confirm what can be licensed, then review a bounded sample under an NDA through a transfer path you approve.
Rights position and sample assessment
03
Dataset and buyer definition
We define what an accepted trace is, which outcomes prove it, and which model builders it suits.
Written dataset specification
04
Terms and delivery
Commercial structure, permitted uses, exclusions, and payouts are agreed in writing before anything is delivered.
Signed agreement and delivery record
Questions
Questions founders ask first
What is a trace?
A record of an AI agent doing a job: the steps, the mistakes, the recovery, and the result.
We are a young company. Do we qualify?
Possibly. Volume, depth, and verified outcomes matter more than age.
What does a trace sell for?
It depends on depth, proof, rights, and buyer demand. We price accepted traces after review. We do not publish a rate.
Our terms let us train our own models. Is that enough?
Not always. Licensing to another company usually needs broader rights. We check this first.
Do you need our code, logs, or production access?
Not to start. The first conversation is business context only. A sample, if needed, comes later under an NDA through a transfer path we approve together.
Start here
Find out what your traces could become.
Tell us what your product does, what it records, and what happens after the agent finishes. Business context only.