Future PresentLabs
Get started

Our open-source mission

A shared foundation.
For everyone.

AI has changed what it means to build software. We believe the old bargain—pay for access to code you cannot understand, change or own—is becoming obsolete.

Explore our GitHub ↗

The value is what we build together.

Before there was a cloud, there were people sharing tools. A compiler. A kernel. A library that solved one difficult problem well. Someone wrote it, someone improved it, and someone else built something neither of them had imagined.

That is still how the world works. Beneath the polished interfaces and billion-dollar platforms are shared foundations—and the people who keep them standing. Open source is not a footnote to modern technology. It is where much of its possibility begins.

A precarious stack representing modern digital infrastructure rests on a small project maintained by one person.
Dependency by Randall Munroe, XKCD. Unmodified artwork, CC BY-NC 2.5. Reuse terms.

AI makes it easier to write software. It does not make the work of maintaining, understanding and trusting it disappear. We want to use that new capacity to strengthen the commons, not build another wall around it.

We open source the work we believe will provide value to our customers and users. Our ambition is a shared foundation that anyone can inspect, improve and build on—from independent developers to teams operating their own infrastructure.

That is why governance belongs in the open core. Policy, grants, distribution, node metrics and audit interfaces should not require a subscription. Your ability to operate a system should not disappear when you stop paying us.

Build with us

Contribution is more than code.

Our contribution is the engineering behind independently usable tools: infrastructure, inference, protocols and the connections between them.

Help make them better with a reproducible bug report, a test, clearer documentation, a driver or an integration. Share what breaks, what works and what your users need.

Find projects and contribution guides →

Your contribution

Start with the repository’s README, license and contribution policy. Check whether a CLA is required before submitting.

Find your starting point →

Different licenses. One purpose.

MIT & Apache 2.0: make the foundations easy to build on.

Permissive licenses let people use, modify and redistribute software, including within commercial and proprietary products, subject to their terms. MIT keeps the requirements simple; Apache 2.0 also provides an explicit contributor patent grant and additional notice requirements.

Mycelium declares MIT OR Apache-2.0. UMIE declares Apache-2.0. The applicable repository and release remain the source of truth.

MIT license · Apache 2.0 license

AGPLv3: keep improvements part of the commons.

For some products, we choose strong copyleft. AGPLv3 protects the shared foundation by requiring corresponding source under its terms when covered software is distributed, and by requiring operators of modified versions to offer corresponding source to users interacting with them over a network.

It permits commercial use. It does not mean every service next to it must become open source; obligations depend on how the covered work is modified, combined and offered. Legion-of-BOM uses AGPL-3.0-or-later.

Read AGPLv3

Contributions that can keep moving.

Good tools outlive their first authors. They move between teams, grow in unexpected directions and become foundations for work no one could have planned. We want the things we build together to have that kind of future.

Where a project uses a Contributor License Agreement, it makes the rights to a contribution explicit—including alternative licensing where the agreement provides for it. The public license remains your license to use the project. A CLA is not a subscription or a paywall.

Each repository publishes its own license and contribution policy. The terms should be clear before you contribute, and the shared foundation should remain something you can build on.

Open-weight models. Intelligence you can operate.

We want useful intelligence to be deployable on infrastructure you control, not available only through a remote API. Open-weight releases can make that possible.

But open weights are not automatically open-source AI. Model weights, inference code, training code and data information are distinct artifacts, with distinct permissions. Each release should state what is available, what you may do with it, and what remains restricted.

Our direction is to release useful models with clear licenses and supporting documentation. That is not a claim that every model is already published: Verk’s research methods and datasets remain private today, and any model release needs its own explicit terms.

Why the distinction matters

Pay us for making it work. Not for keeping it closed.

Managed hosting, dependable operation, integration and enterprise support are services worth paying for. They save time and make shared foundations easier to use. Self-hosting remains a real choice.

Explore the projects · Build with FPL Cloud · Talk about support