Data Partnerships
Data partnerships

Real work leaves a record.

Companies create useful records every day: code, documents, project histories, messages and financial records. Together, they show how work gets done.

Patronus AI licenses company records and private codebases to build realistic environments and training material for AI. We work directly with the people who own the data and with partners who can introduce us to them.

Tell us about a company

Start with what you know.

Tell us a little about the company, the software it uses and the records or code it may be able to share. You don’t need to have every answer. You can submit your own company or introduce one you know. We’ll speak with the owner before taking anything further.

Get started →
✓ You won’t be asked to upload files at this stage.
01
Company
Tell us what the company does and, if you know, its approximate size.
02
Software
Select the tools the company uses. Add what you know and skip anything you don’t.
03
Records and context
Tell us which records or code may be available, how much history exists and how you’re connected to the company. If there’s enough information, we’ll show you an indicative range.
How we handle data

Clear terms before anything is shared.

Company records and code can contain sensitive information. We agree on how material will be reviewed and handled before asking for a sample.

01 / Before review

Agree on evaluation terms

We don’t request files through this form. If an opportunity looks like a fit, we agree on evaluation terms with the owner before receiving a sample. Those terms cover who can review it, where it will be reviewed and what happens if we don’t proceed. See how we handle a sample.

02 / During review

Review the material

Our team reviews the sample to understand what’s available and whether it meets our needs. The sample is used for evaluation, not training.

03 / Before use

Prepare and validate the data

If we move forward, Patronus processes the licensed material and checks it for personal information before production use. We provide a report and handle deletion according to the agreed terms.

From introduction to licence

How it works

Every opportunity starts with a conversation. The owner decides whether to proceed at each stage.

01

Tell us about the company

Share what you know about the company and the records or code that may be available. You can submit your own company or introduce one as a data partner.

02

We speak with the owner

We confirm what exists, who owns it and whether the owner is interested in exploring a licence.

03

Agree on evaluation terms

If there’s a potential fit, we agree on terms before any sample is shared.

04

Review a sample

We review a sample to understand the material, its history and how the pieces connect. For codebases, we also look at the development history and supporting material where available.

05

Agree on a licence

If we want to move forward, we agree on the scope, permitted uses, price, payment terms and acceptance criteria with the owner.

06

Complete the review

We evaluate the full dataset against the agreed criteria and prepare it for use. Payment follows the terms of the agreement.

Questions

A few things you might want to know.

How do you calculate the estimate?

For codebases, we consider the size of the code, its development history and supporting material such as pull requests and tests: $100 to $200 per million tokens of source plus git history, with a rare language or an ageing enterprise system worth one and a half times, and documents that explain the code paid separately per accepted artifact. For company records, we consider the company’s size, the amount of history available and the systems the records come from.

The initial range is an estimate. We review a sample before making an offer, and the full dataset is assessed against the agreed acceptance criteria.

Why didn’t I receive an estimate?

We need a few details before we can provide a useful range, including the company’s size, how much history it has and which records may be available. If you don’t know those details, you can still tell us about the company. We can work through them with the owner.

Do I need to know all the software a company uses?

No. Share what you know and skip the rest. We can fill in the details later.

What makes a codebase or set of records useful?

We look for material that helps us understand how work happened over time. For company records, that might include documents, messages, decisions and the work that followed. For code, it can include the source code, development history, pull requests, issues and other supporting material.

What happens if I connect GitHub?

If you choose to connect GitHub, we review the repositories you select to understand their size, languages and development history. This can include commits, merged pull requests, contributors, tests and CI. We measure them from a temporary copy that is deleted afterwards; no code is kept. The connection is read-only, and you can disconnect it at any time.

What happens to personal information?

We don’t ask you to upload records through the initial form. If we request a sample, we do so under agreed evaluation terms. If we license the material, we process it and validate it for personal information before production use. We discuss reporting and retention as part of the agreement.

Who are data partners?

Data partners are people who can introduce us to companies or founders who may have relevant records or code. They include advisers, accountants, operators, investors, researchers and others who know or can find companies. Partners make the introduction. We work directly with the owner on the review and any agreement. Become a data partner →

How are data partners paid?

Partners can receive a payment for a qualified opportunity and a success fee when a licence is completed. We explain the criteria and agree on payment terms in writing.