Commercial licensing
The code is open source and the data is licensed.
PIT ships two different things under two different licenses, and they do not overlap. The evaluation harness and the tools around it are open source: run them, read them, fork them, ship them inside a product you sell, and you owe us nothing. The rows the harness reads are ours to license. A self-serve plan covers work you run for yourself or for your employer; once your own users receive PIT data, or you pass rows on to anyone outside your company, you need a commercial license from AQX LLC. Those are priced by agreement, so they start with an email rather than a checkout.
The two licenses
Which license covers which piece
The split follows what each piece is. Software we wrote is under a standard open license and stays there. The corpus is a record we pay to keep on a clock, so it is the part that carries terms.
| Piece | License | What that means |
|---|---|---|
The evaluation harness, its gate, its replay, and certverify |
Apache-2.0 | Run it, read it, modify it, redistribute it, and ship it in something you sell. No account and no permission from us. |
| The corpus: API answers, flat files, benchmark releases, the live tape | Your PIT plan | Self-serve covers internal research and apps you operate for yourself or your employer. Customer-facing or redistributed use needs a commercial license. |
| A certificate your own run produced | Yours | Publish it wherever you like. It carries hashes, arm deltas and quotes from your agent’s own rationale rather than PIT rows. |
| Your agent, your prompts, your model keys | Yours | The harness runs on your machine with the network off, so none of it reaches us at all. |
The two never contradict each other because they cover different objects. An Apache-2.0 runner reading licensed rows is the ordinary arrangement for any client library that talks to a paid API, and the harness is that client library with a receipt attached.
Where the line falls
Who ends up holding the data
Checkout asks you to tick one attestation. It reads: “I will use PIT for internal research and applications I operate for myself or my employer. If my product’s end users receive this data, or I redistribute it, I need a Commercial License.”
| Use | License |
|---|---|
| Internal research, and apps you operate for yourself or your employer | Self-serve (Researcher, Power or Desk) |
| Running the open-source harness, on any corpus you are licensed for | Apache-2.0, no plan needed for the code |
| Publishing a certificate, or listing one in the registry | Included with any plan |
| Your end users receive PIT data | Commercial |
| You redistribute, resell, or sublicense rows | Commercial |
| PIT data is embedded in a customer-facing product | Commercial |
| Training a public model for other people to use | Commercial |
| Building a competing information API | Unlicensed |
Three things sit outside any data license: a competing product, resale of the corpus, and a derived archive of publisher bodies presented as an official record. For everything else the terms follow from volume and audience. None of the three restricts the harness, which stays Apache-2.0 whatever you build with it.
Getting in touch
What to put in the email
Say who your end users are, whether you would pass on rows or only features derived from them, and roughly what volume you expect. Terms follow from audience and volume, and the reply quotes a price. The public sample endpoints stay open while we work it out, so your evaluation runs in parallel.