ThinkFacility

News

Intrinsic open-sourced the robot control stack it uses in real manufacturing deployments, under Apache 2.0

The GitHub repository carries one line under the license: this is not an officially supported Google product.

Intrinsic put the robotics stack it runs day to day on GitHub, under Apache 2.0, and announced it at ROSCon 2026 in Toronto. The release is called Intrinsic Core, and the company says it's the same capabilities and services it uses for real manufacturing deployments.

The idea is that you stop writing the boring half. Intrinsic describes pre-made building blocks of robotic behavior you combine, rather than coding robotic capabilities or infrastructure from scratch.

Release
Intrinsic Core, announced at ROSCon 2026 in Toronto
License
Apache 2.0
Where
GitHub, intrinsic-ai/intrinsic-core
The repository this morning
54 Commits, 14 stars, 0 forks
ROS 2 distribution
ROS 2 Lyrical Luth
Also released
the Open Machine Tending Solution (OMTS)

What's in it

Most of it is in the control framework. Intrinsic Control runs a deterministic loop that can switch controllers inside a single cycle off live sensor data, so a robot changes what it's doing mid-trajectory when something in front of it moves. It's hardware-agnostic too, and that's the part that saves days.

This lets you swap out different robot arms, grippers, sensors, and more without needing to rewrite drivers

From Introducing Intrinsic Core™ : An open source approach to Physical AI | Intrinsic

The rest covers grasp planning, motion planning that auto-generates collision-free paths, simulation powered by Gazebo, camera calibration, and pre-configured ROS drivers for supported robots, grippers and 3D cameras. Pose estimation uses NVIDIA FoundationPose for six degrees of freedom on 3D parts (Intrinsic says that removes the need for rigid, expensive physical fixtures, which in a machine shop is real money).

Intrinsic shipped a worked example as well. The Open Machine Tending Solution is a reference design for a CNC machine shop, built to be customized, and it takes hardware from FANUC and Universal Robots.

The line under the license

The README's last words before the trademark notice surprised me.

This is not an officially supported Google product

GitHub - intrinsic-ai/intrinsic-core: Intrinsic Core™ provides an open, local runtime, SDK, and hardware agnostic, real-time control framework for industrial robotics. · GitHub

Contributions need a signed Contributor License Agreement and C++20, and the repository had 54 commits, 14 stars and no forks when we read it this morning. Intrinsic is straight about the limits, though: using Intrinsic Core still takes some basic proficiency in robotics, and the company calls this the early days of its open source journey.

Who it's aimed at

Intrinsic's own view of the market is in the announcement. Tens of thousands of businesses fabricate parts in the US and Europe, only a small fraction of them have any form of automation, and the reason the company gives is supply.

There’s still low to no supply of affordable, manageable AI robotic solutions for them

From Introducing Intrinsic Core™ : An open source approach to Physical AI | Intrinsic

The AI for Industry Challenge that Intrinsic ran with Open Robotics on dexterous cable routing drew more than 5,000 developers and roboticists across 115 countries. CEO Wendy Tan White announces the top five winners on Wednesday, September 23 at 5:20pm, and a panel on the challenge follows on Thursday, September 24 at 5:40pm.