Skip to content
Tool

Solarian

Open-source Python framework for building, teleoperating and training robots

Solarian is an open-source AI tool for developer tools. It is available on CLI, Linux, macOS, Windows, Self-hosted and API.

Pricing
Open source
Platforms
CLI, Linux, macOS
Listing updated

Similar tools

About Solarian

The README-style site positions Solarian as a Python framework that covers the whole robot stack, from hardware drivers to AI control. A robot is assembled from separate modules such as an arm, gripper or camera, and each module runs in its own process at its own rate, so a fast actuator does not wait on a slower camera.

Control is split across a Robot, a relay Server and a Controller, connected over ZeroMQ. An AI Agentic Controller wraps a vision-language model so that a task given in chat is planned and then executed through the robot's physical skills exposed as tools. Setting a flag switches the same module definitions and YAML from hardware to a MuJoCo simulation.

The framework targets robotics developers and researchers who want to move from teleoperated demonstrations to trained policies in one toolchain. Extension happens through installable hardware and skill plugins, and the project is configured with Hydra YAML files.

What it does

Modular multi-process control
Each robot module runs in its own process at its own rate, coordinated over a shared control loop.
Built-in inverse kinematics
The module solves end-effector commands each tick and reports the live pose via forward kinematics.
AI agentic controller
Drives skills as tools from a chat task using any vision-language model.
Simulation switch
Setting is_sim to true runs the same modules and YAML in a MuJoCo world instead of on hardware.
Plugin system
Hardware and skill plugins install with a command, and a scaffold command creates new ones.
Session recording and datasets
Records camera video and signals at native rate and supports splitting, trimming and merging episodes.
Policy training and serving
Trains diffusion or ACT policies locally, on Modal or on Solarian Base GPUs, then serves them to skills.

Good for

  • Prototyping control software for a custom arm or gripper before buying full hardware
  • Collecting teleoperated demonstrations and training a policy for a manipulation task
  • Testing robot behaviour in MuJoCo simulation before deploying to a physical robot
  • Letting a robot be commanded from a chat interface using a vision-language model

Pricing

Open source

Open-source framework; Solarian Base GPUs are referenced for training but no prices are given.

Prices change. Check solarian.dev before you buy.

Solarian FAQ

Is Solarian open source?

The site describes Solarian as an open-source framework. It does not name a specific licence on the pages provided.

Which learning methods does Solarian support for training policies?

The site lists diffusion and ACT policies, and says you can also bring your own trainer.

How are hardware and skills added?

Hardware and skills are added as plugins with the solarian install command, and solarian create-plugin starts a new one.

Can I run a robot without hardware?

Yes. Setting is_sim to true runs the same modules and YAML in a MuJoCo simulation.

Is Solarian free?

Yes. Solarian is open source, so you can use and self-host it for free.

What are the best alternatives to Solarian?

The closest alternatives to Solarian are Lynx, CopilotKit, AiSixteen, Buildt and Github.

Where this comes from

  • Description: written from the project's website.