Installation¶
Choose a deployment pattern¶
OpenCastor supports four deployment patterns from the OpenCastor stack spec:
- Declaration-only.
pip install robot-md. Manifest authoring, schema validation, dry-run agent planning. No physical robot moves. - Gateway Mode.
pip install robot-md-gateway+ agateway initwizard. Single-robot local deployment with the open safety kernel. - OpenCastor Full (this guide below). The productized runtime — gateway embedded, plus drivers, fleet, cloud bridge, UI.
- Regulated overlay. Any of the above + signed evidence bundle export to RRF + jurisdiction-specific compliance packets.
This installation guide covers Pattern 3.
Requirements¶
- Python 3.10+
- Any Linux/macOS board with 1 GB+ RAM (Raspberry Pi 4/5, Jetson, x86)
- Optional: Hailo-8L NPU, Coral TPU, CUDA GPU
Install¶
bash
pip install opencastor
bash
git clone https://github.com/craigm26/OpenCastor.git
cd OpenCastor
pip install -e .
First boot¶
bash
castor init # interactive setup wizard
castor status # verify runtime is healthy
castor gateway start # start the local HTTP gateway
The gateway runs on http://localhost:18789 by default.
Verify¶
bash
curl http://localhost:18789/api/status
Expected response:
json
{
"version": "v2026.3.30.0",
"rcan_version": "2.2.0",
"runtime": "ok",
"p66": "active"
}
Next steps¶
- Quickstart → — connect a robot and run your first command
- Configuration → — tune the runtime for your hardware
- Hardware Profiles → — Pi 5 + Hailo, Jetson Orin, CPU-only setups