Intrinsic Open-Sources the Robotics Stack It Uses in Real Factories, Under Apache 2.0

The Release

Intrinsic, an Alphabet company spun out of Google X in 2021, published Intrinsic Core to GitHub on September 22, 2026, at ROSCon 2026 in Toronto. The Apache 2.0 license carries no restrictions on commercial use, sublicensing, or redistribution.

Intrinsic runs this exact software in its own real-world manufacturing deployments. That provenance matters: the platform came from factory floors, with production-grade history behind each module.

What's in the Stack

Intrinsic Core ships a set of integrated modules. The centerpiece is Intrinsic Control, also called ICON, a hardware-agnostic real-time control framework that can switch between controllers mid-trajectory based on live sensor data. Developers can swap robot arms, grippers, and cameras without rewriting hardware drivers.

Pose estimation runs on NVIDIA's FoundationPose model, giving robots 6-degree-of-freedom object detection without rigid physical fixtures. Motion planning auto-generates collision-free paths, replacing the manual joint-by-joint programming common in traditional setups.

Grasp planning uses real-time sensor feedback so grippers adapt dynamically to a part's location and orientation as it arrives. Simulation runs on Gazebo, with automated camera calibration and pre-configured ROS 2 drivers completing the package.

First Reference Application

Alongside Intrinsic Core, Intrinsic released the Open Machine Tending Solution, a customizable reference design for AI-enabled CNC machine tending. It supports FANUC and Universal Robots hardware out of the box. The broader hardware ecosystem also includes ATI, Schunk, Robotiq, and Basler.

Intrinsic describes the machine tending solution as a starting point for systems integrators. The base layer is there; integrators build the application on top of it.

The Access Problem

Tens of thousands of businesses fabricate parts across the U.S. and Europe, and by Intrinsic's own count only about 8% run any automation. The company frames the release as a direct response: give developers a working starting point so they spend more time solving the manufacturing problem itself instead of rebuilding the same foundations from scratch.

Brian Gerkey, CTO at Intrinsic, called it "a durable commitment from Intrinsic to share this with the community." Geoffrey Biggs, CTO of the Open Source Robotics Foundation, called the release "both the first step and a demonstration of where I think robotics is going to go."

Community Signal

Before the public launch, Intrinsic ran an AI for Industry Challenge to stress-test the stack in competitive conditions. More than 5,000 developers and roboticists from 115 countries entered; 2 teams achieved 100% success rates on the final electronics assembly task, identifying cable connectors and inserting them into the right ports on a live taskboard.

Intrinsic Core targets Ubuntu 22.04 LTS, 24.04 LTS, or 26.04 LTS and builds around the ROS 2 Lyrical Luth distribution. The repository is live at github.com/intrinsic-ai/intrinsic-core.

Sources

• Introducing Intrinsic Core: An open source approach to Physical AI, Intrinsic Blog
• Google's robotics unit Intrinsic open-sources its foundational infrastructure for intelligent robots, SiliconAngle
• Intrinsic open sources key parts of its platform for easier development, The Robot Report
• Intrinsic Open-Sources Core Robotics Capabilities at ROSCon 2026, Unite.AI
• Intrinsic open sources core industrial robotics technology, Robotics and Automation News
• Intrinsic open-sourced the robot control stack it uses in real manufacturing deployments, under Apache 2.0, Think Facility

Companies mentioned

• Intrinsic
• NVIDIA
• Gazebo
• FANUC
• Universal Robots
• Open Source Robotics Foundation