Intrinsic published Intrinsic Core on GitHub on September 22, at the ROSCon conference in Toronto. The free toolkit covers real-time control, finding parts with cameras, motion and grasp planning, and simulation for robot arms. Its Flowstate tool, advanced AI models, and cloud services remain paid.
New to this? Read it in simple words
- On September 22, Google’s Intrinsic released Intrinsic Core, a free toolkit for robot arms. It covers real-time control, finding parts with cameras, planning movements and simulation.
- The code is open source. But Intrinsic’s Flowstate tool, advanced AI models and cloud services still cost money.
- Free building blocks can make factory robot projects cheaper to start. Intrinsic says users still need basic robotics skills.
- Intrinsic says it uses the same software in real projects, but named no customers. We found no independent tests yet.
- Open source
- Software whose code anyone can read, use and change.
- Simulation
- Testing how something works on a computer instead of in the real world.
What is in the free toolkit
Intrinsic started in 2021 as a project of Alphabet, Google’s parent company, and became part of Google in February 2026. It announced Intrinsic Core at ROSCon in Toronto, a conference for the Robot Operating System, or ROS.
The code is on GitHub under the Apache 2.0 licence. Companies may use and change it, including in products that they sell.
Core includes real-time control that can change a robot’s path during a move, based on sensor data. It can find parts with cameras, using NVIDIA’s FoundationPose, so parts do not need fixed holders. It also plans collision-free movements and grasps, and it includes simulation, camera calibration, and drivers.
Intrinsic also released OMTS, an open example application for loading and unloading CNC machines.
Core runs on a local x86 PC. Camera features need an NVIDIA graphics card and NVIDIA’s separate model licence.
What users need, and what still costs money
Everything runs on the user’s own computer. The setup guide asks for an x86 PC with at least 32 GB of memory. The camera features need an NVIDIA graphics card, and the first build takes about 50 minutes.
The Apache licence covers Intrinsic’s code, but not the FoundationPose model itself. Users download that model from NVIDIA under NVIDIA’s own licence.
Flowstate, Intrinsic’s development tool, is not part of the release. Neither are its more advanced AI models or its industrial cloud services. Intrinsic calls them enterprise services that work with Core.
Free basics, paid extras
Forbes compares the approach with Android. Google gives away the basic layer and charges for the AI models and cloud services on top.
Brian Gerkey, Intrinsic’s chief technology officer, told The Robot Report that “what we’re sharing this week is the core of our own software stack”. He also chairs the board of Open Robotics, the non-profit that looks after ROS.
Outside contributors must sign Google’s contributor agreement, so Intrinsic controls the project. Intrinsic says users still need basic robotics skills.
Intrinsic says Core is the same software it uses in real deployments, but it named no customers that use it. We found no independent tests yet.
Sources
Every fact in this story comes from the sources below. Open them to check our work.
- 1Primary source · September 22, 2026Introducing Intrinsic Core™: An open source approach to Physical AI Intrinsic
- 2
- 3Research · September 22, 2026Intrinsic open sources key parts of its platform for easier development The Robot Report
- 4
- 5Research · September 22, 2026Google’s robotics unit Intrinsic open-sources its foundational infrastructure for intelligent robots SiliconANGLE
- 6
We read Intrinsic’s announcement and its GitHub repository, then compared them with The Robot Report, SiliconANGLE, and Forbes. Hardware needs come from the repository’s own guide. The claim that Core is the software Intrinsic uses in real deployments comes from the company.