CenterOS
The operating system for robotics AI development. CLI tools, MCP server, and 39 platform actions -- all from your editor.
Three Ways to Build
CenterOS CLI
pip install centeros-cliLogin, manage datasets, launch training, deploy models -- all from your terminal. Rich tables, formatted output, and Python SDK access.
$ centeros login $ centeros datasets list $ centeros training create
CenterOS MCP
pip install centeros-mcpFor Claude Code, Cursor, and Windsurf. 43 tools, 55 resources, and 8 code gen prompts. Your AI auto-references wiki docs when writing robot code.
- 43 platform tools
- 55 live wiki resources
- 8 code generation prompts
REST API
39 actionsDirect HTTP access to every platform action. Discover with GET /api/actions, execute with POST /api/actions/{name}. PAT authentication with frl_ tokens.
$ curl /api/actions \ -H "Authorization: Bearer frl_..."
Quick Start
Get up and running in three steps.
Install the CLI and register
$ pip install centeros-cli $ centeros register # or: centeros login
Install the MCP server
$ pip install centeros-mcp
Add to your editor's MCP config (e.g.
~/.claude/.mcp.json)
{
"mcpServers": {
"centeros": {
"command": "centeros-mcp",
"args": ["--token", "frl_YOUR_TOKEN"],
"env": {
"CENTEROS_API_URL": "https://platform.roboticscenter.ai"
}
}
}
}
What CenterOS MCP Gives Your AI
When you connect the MCP server, your AI assistant gains four layers of robotics intelligence:
| Layer | Count | What |
|---|---|---|
| Platform Tools | 39 | Datasets, training, models, fleet, simulation, processing |
| Wiki Knowledge | 33 | Live docs from roboticscenter.ai/wiki -- hardware guides, SDK reference, troubleshooting |
| Hardware Specs | 17 | Wuji Hand, OpenArm, BrainCo Revo, VLAI L1, Booster K1, Damiao AGV, and more |
| Code Gen Prompts | 8 | new_teleop_agent, hardware_setup, train_and_deploy, data_pipeline, and more |
The AI automatically searches wiki documentation and hardware specs when generating code. No manual lookup needed -- ask in natural language and it finds the right reference.
Example Interactions
Here is what working with CenterOS MCP looks like in practice:
datasets.upload to get a signed URL, uploads your data, then calls training.create with ACT config, monitors job status, and reports back when training completes.ip link set can0 up type can bitrate 1000000 with LED status verification.Supported Hardware
CenterOS MCP includes specs, setup guides, and troubleshooting for 17 devices across 5 categories: