Video link
https://www.loom.com/share/77bf2fe9627549608e6b6d7ae38da915?sid=86324aaa-0a25-4b8b-a5f6-4e7217a275bc
Inspiration
We wanted to create a solution that empowers anyone, regardless of technical background, to bring robotic ideas to life from their browser.
What it does
BotBench is a web-based ROS playground that lets users write scripts, generate code using natural language, simulate robots in 3D, and get real-world hardware recommendations—all without any local setup. Makes it easier for the users to easily scale up the scripts to professional simulators.
How we built it
We built BotBench using Next.js Vercel v0. For AI-driven script generation, we integrated Anthropic MCP, Gumloop MCP. The 3D simulation viewport uses Three.js. The scripts generated can be imported and refined further in Gazebo or Unity. Hardware recommendation engine is built using Perplexity MCP, Firecrawl MCP.
Challenges we ran into
- Building the web-based simulation for the generated scripts
- Building reliable AI-driven code generation without creating safety or logic errors in the simulation
Accomplishments that we're proud of
- A functioning in-browser ROS coding and simulation environment
- Generated and ran natural language-driven robot behaviors
- Designed a hardware recommendation engine to bridge simulation with real-world builds
What we learned
- The ease with which MCP enables integration with multiple APIs, allowing quick prototyping and development of modern AI applications
- Challenges in the robotics sim to real problem
What's next for BotBench
- Add support for drag-and-drop block-based coding for K-12 learners
- Generating actual code that can be uploaded on the physical hardware
- Integrations with multiple physics engines out of the box
- Creating a marketplace for robotics projects presets
Built With
- anthropic
- gumloop
- next.js
- perplexity
- python
- v0
- vercel
Log in or sign up for Devpost to join the conversation.