OCUDU AI Agent Skills¶

Published: June 23, 2026

Agent skills for monitoring and operating an Ocudu 5G RAN stack using AI coding agents such as Claude, Cursor, and Windsurf. Built on the open Agent Skills standard.

Note: This documentation is a replica of the README available at the github/ngkore/OCUDU-AI-Agent-Skills. Please refer to the original repository for the most up-to-date information.

Repository Structure¶

.agents/
└── skills/
    └── ocudu/
        ├── SKILL.md              # Skill instructions read by the agent
        ├── scripts/
        │   └── ws_metrics.py     # WebSocket metrics collector
        └── references/
            ├── metrics_schema.md # JSON metrics field reference
            └── config_params.md  # gNB YAML config reference

Setup¶

1. Install the Skill¶

Clone the repository into your project or copy the .agents/ directory:

git clone https://github.com/ngkore/OCUDU-AI-Agent-Skills.git

2. Install Dependencies¶

pip install websocket-client

3. Configure the gNB¶

Enable remote_control and metrics.enable_json in the gNB YAML configuration:

metrics:
  enable_json: true
  layers:
    enable_sched: true
    enable_mac: true
    enable_rlc: true
    enable_ngap: true
    enable_rrc: true
    enable_executor: true
    enable_app_usage: true
    enable_du_low: true
    enable_ru: true

remote_control:
  enabled: true
  bind_addr: 127.0.0.1
  port: 8001

Use bind_addr: 0.0.0.0 if connecting from a remote machine.

Usage¶

Once the skill is in place, ask your AI agent natural language questions about the running gNB:

> How many UEs are connected and what's their signal quality?
> Are there any scheduling issues or HARQ failures?
> What's the CPU and memory usage of the gNB?
> Collect metrics during a ping test and summarize the results.
> Is the AMF connected? How many PDU sessions are active?
> What MCS and modulation scheme is the UE using?
> Check if the gNB is running.
> Run a ping test from the UE to the core network.

How It Works¶

The agent connects to the Ocudu gNB WebSocket interface (default 127.0.0.1:8001), subscribes to JSON metrics, and interprets the data to answer questions. Ocudu streams metrics over WebSocket — no REST API is involved.

Requirements¶

Ocudu gNB with remote control enabled
Python 3.8+
websocket-client package

Testing the Metrics Collector¶

Run the metrics collector manually to verify connectivity:

pip install websocket-client
python3 .agents/skills/ocudu/scripts/ws_metrics.py --url 127.0.0.1:8001 --count 5 --pretty

Tutorial Video¶

OCUDU AI Agent Skills Tutorial