Network Management Blog - by StableNet^®
Regular posts on all things around automated network & service management

Features, trends and new product development

StableNet^® & Agentic AI: Introduction & Basic Setup

May 29^th 2026, Würzburg

This blog post explores how StableNet^® can be extended with Agentic AI to enable natural language interaction with network data and operations.

The focus is on a secure, local setup using a Large Language Model (LLM) to automate typical tasks such as health checks, alarm analysis, and reporting. The goal is to bridge the gap between the StableNet^® REST API and an intelligent assistant that can interpret requests and act on them.

Prefer to see it in action? Visit our YouTube channel to watch the full StableNet^® & Agentic AI Snapshot.

The Idea: Natural Language Instead of Manual Operations

Working with network management systems often involves navigating complex interfaces or issuing repetitive API calls. By introducing an AI agent, this interaction model can be simplified significantly.

Instead of manually querying data, operators can issue requests in natural language, for example:

Check the current system status
List open alarms and group them by severity
Generate a summary for management

The agent translates these requests into API calls, processes the responses, and returns structured results.

This approach helps to:

reduce manual effort
speed up analysis and troubleshooting
improve overall operational efficiency

Technical Setup Overview

The setup is designed to run fully locally to ensure that sensitive data remains within the environment.

Local LLM

A locally hosted model runs directly on the host system. This avoids external dependencies and ensures that no data leaves the environment.

Sandboxed Agent

The Agentic AI runs inside a virtual machine (e.g. Ubuntu). It is fully isolated and has no direct access to the network or external systems.

Proxy Server

A lightweight Python-based proxy acts as an intermediary between the agent and StableNet^®. All API communication is routed through this component.

Security Considerations

A key aspect of the setup is strict control over what the AI can and cannot do.

The agent does not have access to credentials
All authentication is handled by the proxy
Access can be restricted to read-only operations
Requests are rate-limited to prevent misuse
All interactions are logged for traceability
No infrastructure discovery (e.g. scanning) is possible

This ensures that even in the case of unexpected behavior, the system remains controlled and predictable.

Configuring the Agent

Environmental monitoring should not be seen as a standalone solution. It should be part of a complete data center monitoring approach. A centralized view of all data center components on a single platform enables comparisons to be made and allows the performance of a data center to be viewed holistically.

Let’s look at a few key points that a solution should cover.

Example Use Cases

Once configured, the agent can be used for a variety of operational tasks.

Checking Key Endpoints

The agent can query multiple StableNet^® endpoints and provide a consolidated overview of their status. It can identify which endpoints respond correctly and which are not accessible under current permissions.

Alarm Analysis

Beyond simple filtering, the agent can interpret alarm data semantically. It can identify specific alarms, correlate them with devices, and extract relevant context from the API responses.

Management Reporting

The agent can generate structured summaries based on current data. This includes:

grouping alarms by severity
identifying affected systems
providing short recommendations

This reduces the effort required to prepare regular reports.

Automation

Executed workflows can be turned into reusable scripts. The agent can generate shell scripts and suggest scheduling (e.g. via cron), enabling recurring tasks such as periodic health checks or reporting.

Conclusion

By introducing a proxy layer and a lightweight configuration for the agent, StableNet^® can be extended with an intelligent interface that simplifies access to network data.

The result is a controlled, local solution that: