Want to connect with Innate?
Join organizations building the agentic web. Get introductions, share updates, and shape the future of .agent.
Is this your company?
Claim this profile to update your info, add products, and connect with the community.
Innate is a bridge between digital agents and the physical world. While much of the AI agent ecosystem focuses on software-to-software interactions, Innate provides the "embodied" layer of the stack. They are building a platform where an agentic operating system (BASIC) manages the reasoning and execution of physical tasks, effectively turning a robot into a hardware manifestation of a reasoning agent.
For developers in the agent community, Innate is relevant because it offers a standardized, programmable target for foundation models that can act in the real world. By treating physical actions as skills that a central agent can orchestrate, they apply the same architectural patterns seen in browser-based agents—like goal decomposition and tool use—to robotic manipulation. This makes them a primary player in the "physical agent" or "embodied AI" category, providing a path for agentic software to escape the screen.
Innate is a Silicon Valley startup attempting to shift robotics from the factory floor to the living room. Founded by Axel Peytavin and Vignesh Anand—engineers with backgrounds at 1X Technologies and various top-tier AI labs—the company is building Mars, a general-purpose robot designed to be owned, programmed, and modified by individual users. The company is backed by Y Combinator and a group of researchers from major tech labs, including Google and Meta.
The philosophy behind Innate is a rejection of the centralized, uniform future often depicted in science fiction. Instead of a single corporation controlling a fleet of identical machines, Innate advocates for an open platform where developers build and share their own behaviors. They cite the early days of Silicon Valley hackers as their inspiration, aiming to provide the hardware foundation that allows physical automation to become as versatile as the PC.
The Mars robot is an open hardware platform equipped with an Nvidia Jetson Orin Nano 8GB GPU, providing enough local compute for real-world reasoning. Its sensor suite is built for spatial awareness, featuring 2D LiDAR, stereo cameras, and an RGB camera on a moving head. The primary manipulator is a 6-DOF (six degrees of freedom) arm designed specifically for reinforcement learning and skill acquisition.
Crucial to the Mars experience is BASIC, which Innate describes as an "agentic OS." This operating system runs what they call a programmable foundation agent. Unlike traditional robotics software that relies on rigid, pre-defined scripts, BASIC is designed to reason over long-term goals and adapt to environmental changes in real time. It treats physical actions as "skills" that the core agent can call upon to complete a task. This architecture allows the robot to handle unexpected interruptions or changes in its surroundings without the entire program failing.
Programming a Mars robot involves writing "behaviors" in a Python-based syntax. A behavior consists of a directive—a high-level prompt that guides the agent's logic—and a list of available digital or physical skills. For example, a chess-playing behavior would include digital skills for computing moves and physical skills for moving pieces on a board. This abstraction allows developers to focus on the logic of the task rather than the low-level mechanics of motor control.
Training the robot is designed to be accessible. Innate provides a system where users can attach their phone to the robot's arm to collect training data via teleoperation. They claim that a new physical skill can be taught in less than 30 minutes of data collection. This low barrier to entry is intended to foster a community of users who create and share behaviors for everything from sorting trash and cleaning floors to acting as a "barman" or a security monitor. By open-sourcing the hardware and OS, Innate is betting that a decentralized ecosystem of tinkerers will discover the most valuable use cases for domestic robotics.
A general-purpose personal robot running an agentic operating system for real-world physical tasks.
A modular high-level library to train embodied AI agents across a variety of tasks and environments.
A flexible, high-performance 3D simulator for Embodied AI research.
Leveraging Large Language Models for Visual Target Navigation
WebSocket gateway for ROS2 topics and services
Library of Behaviors for BASIC
Innate is hiring
You've explored Innate.
Join organizations building the agentic web.