Introduction

Drift is your AI copilot for robotics development. Building, testing, and debugging robot simulations has always required deep expertise across ROS2, Gazebo, URDF, and more- Drift removes that barrier entirely. Describe what you want in plain English, and Drift handles all the complexity behind the scenes. From setting up a workspace to launching a full simulation environment, you just come with prompts, for example:

• create a three arm manipulator with camera sensor
• launch my robot with the custom world file
• add a lidar sensor to my robot and rebuild
• why isn't my camera publishing images?

​

Why Drift?

While tools like Cursor or Claude are excellent for general coding and debugging, they miss context on robotics and that’s what Drift eliminates. In the video below, we tried solving a same Gazebo error using Claude and Drift.

• Claude looks at the surface and concludes Gazebo isn’t installed- It was.
• On the other hand, Drift inspects the actual ROS2 environment: running processes, topic connections, launch file output, and system state. It finds the real cause and fixes it without having to interpret the error.

This is the gap Drift is built for. Not just writing code; understanding the full stack it runs on.

​

How It Works

Drift runs as an interactive CLI agent on your local machine. When you type a prompt, the agent interprets your request, generates the necessary ROS2 packages, URDF/SDF files, and launch files, invokes colcon to build the workspace, and manages Gazebo processes - all without you touching a config file.

1. You describe what you want in plain English at the drift> prompt.
2. Drift plans the steps needed: workspace layout, robot description, world file, launch configuration.
3. Drift executes- running builds, resolving errors automatically, and launching Gazebo.
4. You iterate- add sensors, change configurations, debug issues by asking follow-up questions.

Everything runs locally. Drift does not upload your files or simulation state anywhere.

​

Key Capabilities

Project scaffolding from a single prompt- Describe the robot you need and Drift generates the full project: URDF with correct inertia and collision geometry, ROS2 package structure, Gazebo plugins, controller configs, and launch files. Not templates - working, buildable code from a description. Full simulation stack management- Drift launches Gazebo, spawns your robot, opens RViz with the right config, and bridges all topics in one command. It tracks every background process, so you’re not juggling terminals or hunting for zombie gzserver instances. Debugging and live ROS2 inspection- Describe the symptom and Drift does the detective work: inspects running nodes, checks topic connections and QoS compatibility, validates URDF inertia tensors, and traces the command chain from publisher to controller to joint. For live inspection, it has access to all ROS2 tools - list nodes, echo topics, call services, get and set parameters - all through natural language. Build error recovery- Drift reads colcon output, finds the root cause, and fixes it- whether that’s CMakeLists.txt, package.xml, setup.py, or a missing dependency. When you add a sensor, it updates the URDF, bridge config, and launch file together - changes stay consistent across files. You stay in control- Every command is shown before it runs. Press ESC to pause mid-execution and step through one command at a time, or let Drift run the full task uninterrupted. Use drift -c for CI pipelines and scripted workflows.

​

Who is Drift for?

• Students: learn robotics concepts without getting stuck on setup configurations
• Researchers: iterate faster on algorithms without simulation setup slowing you down
• Engineers: prototype and test production systems in a fraction of the time
• Hobbyists: bring robot ideas to life without needing any deep expertise

​

Current Status

Drift CLI is in open beta (v1.0.15). It runs on Ubuntu 20.04+ with ROS2 and Gazebo Sim. Install Drift or get started with the Quickstart.

​

Known Limitations

Because Drift is in its early versions, there are rough edges worth knowing about upfront:

• Prompt sensitivity: Drift works best with clear, specific descriptions. Very short or ambiguous prompts (e.g., “make a robot”) may produce generic results. More detail yields better output.
• Complex multi-robot setups: Multi-robot coordination and shared worlds are experimental and may require manual adjustments.
• ARM architecture: Drift’s binary targets x86_64 (amd64). Apple Silicon (M1/M2/M3) users can run Drift via VMware Fusion using an x86_64 Ubuntu image. Raspberry Pi is not supported.
• Internet connectivity: Drift requires internet access to process prompts (the AI agent calls an external API). Your local files and simulation stay on your machine.

If you hit an issue, describe it at the drift> with a prompt, the agent will attempt to diagnose and recover. For persistent problems, open an issue on GitHub.

​

Next Steps