Building AI Agents for Network Operations : Design LLM-Powered NetOps Workflows with Python, Ollama, MCP, and Tool Calling.

Format:

Book

Author/Creator:

Baksh, Sif.

Language:

English

Subjects (All):

Artificial intelligence.

Computer networks.

Physical Description:

1 online resource (226 pages)

Edition:

1st ed.

Place of Publication:

Birmingham : Packt Publishing, Limited, 2026.

Summary:

Build AI-assisted network troubleshooting workflows that parse CLI output, call approved tools, use MCP, and keep evidence visible for review Key Features Build local LLM workflows for NetOps using Python, Ollama, and validated CLI data Create troubleshooting agents that use memory, approved tools, and clear evidence Package reusable network.

Contents:

Intro

Building AI Agents for Network Operations

Design LLM-powered NetOps workflows with Python, Ollama, MCP, and tool calling

Contributors

About the author

Table of Contents

Preface

Who this book is for

What this book covers

To get the most out of this book

Download the example code files

Download the color images

Conventions used

Get in touch

Share your thoughts

Free benefits with your book

How to Unlock

Stay Sharp in Cloud and DevOps - Join 44,000+ Subscribers of CloudPro

1

Understanding AI Agents for Network Operations

Technical requirements

The operational problem agents are trying to solve

What is an AI agent?

Chatbots, copilots, automation, and agents

Where agents help in network operations

Alert triage

CLI output parsing

Troubleshooting support

Documentation and handoffs

Where agents are the wrong tool

The basic agent workflow

A practical troubleshooting scenario

Human control is not optional

How traditional automation and agents work together

What you will build in this book

Production reality check

How to think about success

Summary

Get this book's PDF version and more

2

LLM Fundamentals and Local Setup

Understanding why LLM fundamentals matter for NetOps

Understanding what a large language model is

Working with tokens and context windows

Managing output size and cost

Controlling temperature and predictable output

Comparing temperature behavior

Understanding stateless calls and memory

Using Ollama for the local lab

Installing Ollama and pulling the model

Running the model from the command line

Setting up the repository and Python environment

Calling Ollama from Python

Reviewing what the Python code is doing.

Experimenting with temperature from Python

Reviewing tokens in the lab

Troubleshooting and production considerations

Troubleshooting the local setup

Checking production reality

Join us on Discord

3

Prompt Engineering for Network Automation

Understanding why prompt engineering matters

Introducing the RACE prompt framework

Defining the role

Adding anchors

Providing context

Defining the expected output

Comparing vague prompts with structured prompts

Building a configuration parser prompt

Writing the RACE prompt

Running the prompt engineering lab

Reviewing what the lab is doing

Validating model output

Creating reusable prompt templates

Creating an alert triage prompt

Creating a documentation prompt

Handling common prompt failures

Extending prompts to live device data

4

Parsing Network Outputs into Structured Data

Understanding why structured output matters for NetOps

Moving from raw CLI text to validated JSON

Parsing interface output

Reviewing the interface prompt

Parsing BGP summary output

Normalizing multi-vendor output

Validating structured output before using it

Handling malformed output and uncertain states

Connecting parsing to live device data carefully

Choosing between LLM-assisted and deterministic parsing

5

Building a Network Chatbot with Memory

Understanding why chatbot memory matters

Running a stateless chatbot

Adding application-managed memory

Building the NetworkChatbot class

Building the chat method

Building the full prompt

Running the memory-enabled chatbot.

Managing conversation history

Resetting the conversation

Using a network-focused system prompt

Preparing chatbot memory for agents

6

Designing Tools and Agentic Workflows

Understanding why agents need tools

Moving from chatbot responses to tool-assisted reasoning

Reviewing the mock network tools

Reviewing the main Lab 4 script

Mapping tool names to approved Python functions

Describing tools to the model

Building the system prompt for tool use

Parsing model requested tool calls

Executing tools safely

Returning tool results to the model

Limiting the agent loop

Handling tool-calling failure modes

Running the agentic network bot

Trying a multi-step question

Testing an investigation query

Keeping live network access optional

Applying production boundaries to tools using agents

Logging what the agent does

Avoiding common agent design mistakes

Preparing for the main troubleshooting agent

7

Building the Main Network Troubleshooting Agent

Moving from tool mechanics to troubleshooting

Reviewing the troubleshooting scenarios

Reviewing the mock topology

Running the troubleshooting agent

Scenario 1: Checking device status

Scenario 2: Checking BGP health

Scenario 3: Investigating leaf2

Scenario 4: Investigating a missing route to host symptom

Building an evidence-based final answer

Catching wrong or incomplete model conclusions

Improving the troubleshooting prompts

Using the agent interactively

Keeping the workflow mocked and read-only

Understanding what the agent can and cannot do

Walking through the leaf2 investigation step by step

Creating an evidence record.

Checking the answer before trusting it

Designing a structured troubleshooting response

Expanding the missing route investigation

Preparing for reusable tools with MCP

8

From Lab Agents to Reusable Tools with MCP

Moving beyond direct tool calling

Understanding what MCP adds

Reviewing the Lab 5 architecture

Reviewing the safe network tool wrappers

Testing the tool layer before MCP

Exposing network tools through the MCP server

Running the MCP server

Connecting the HTTP bridge

Opening the browser UI

Testing the MCP tools from the UI

Understanding why the bridge exists

Running the lab in the correct order

Troubleshooting the MCP lab

Keeping MCP tools safe

Deciding when MCP is useful

Comparing MCP transport choices

Designing stable MCP tool contracts

Avoiding common MCP design mistakes

Reviewing what we built

9

Moving Toward Production-Ready Network Agents

Understanding why production is different from a lab

Defining the production boundary

Keeping the first operational step read-only

Deciding what should not be automated

Applying authentication and authorization

Handling secrets and credentials

Validating inputs and outputs

Using the Lab 6 safety wrapper demo

Designing tool safety and approval workflows

Logging every tool call

Reviewing an audit event

Observing the agent and tool layers

Handling failure and rollback planning

Using the production skeleton pattern

Planning a staged rollout

Reviewing the production-readiness checklist

Building a production review packet

Testing before a controlled pilot

Writing an operations runbook.

Using feature flags and kill switches

Assigning ownership and support

Defining acceptance criteria

Creating a production-readiness worksheet

Preparing an approval record

Avoiding production shortcuts

Connecting the chapter back to the book's journey

Final go/no-go review

10

Unlock Your Exclusive Benefits

Unlock this Book's Free Benefits in three Easy Steps

Step 1

Step 2

Step 3

Need Help

Appendix A

AI Network Agent Design Toolkit

Book concepts mapped to design artifacts

Use-case fit scorecard

Use-case brief template

RACE prompt worksheet

Reusable RACE prompt skeleton

Structured-output and validation checklist

Minimal schema checklist

Testing different models

Memory and context policy

Tool inventory and safety matrix

Tool contract template

Evidence record for troubleshooting agents

Troubleshooting response template

MCP tool contract worksheet

Production readiness checklist

Audit event template

Read-only pilot acceptance criteria

Operations runbook template

Feature flag and kill switch template

Go/no-go review

Why subscribe?

Other Books You May Enjoy

Packt is searching for authors like you

Index.

Notes:

Description based on publisher supplied metadata and other sources.

Part of the metadata in this record was created by AI, based on the text of the resource.

Other Format:

Print version: Baksh, Sif Building AI Agents for Network Operations

ISBN:

9781808346828

OCLC:

1607915277