CfgMgmtCamp 2026: AI and Automation

CfgMgmtCamp 2026: AI and Automation

This post is part of a series of talks presented at CfgMgmtCamp 2026. Please see CfgMgmtCamp for all other talks.

Asking a local LLM about my Ansible playbooks because why not

Speaker: David Moreau-Simard

Slides: Presentation

Related Pull Request: WIP: First iteration of a MCP server for ara #641

Forum Post: Link

Video: YouTube

LMs can write Ansible roles and playbooks as well as plugins and modules to a relative degree of competency with commodity hardware that you can run offline, at home, with less privacy and security concerns.

They certainly understand what Ansible is, how it works and they can write YAML, Jinja or Python.

What about troubleshooting an issue with an Ansible playbook ?
You could provide it with a log file or copy & paste the results of your ansible-playbook command.
Maybe it could point you in the right direction… or hallucinate. Who knows ? ¯(ツ)/¯

ARA Records Ansible playbook results to a database and provides an API to query the results.
What if we gave a LLM programmatic access to that API using Model Context Protocol (MCP)?
It would allow the model to include context like host facts, playbook files and detailed task results, amongst other things.

The author experimented with it (for science) and the results are interesting.

Join us to learn about the experiment, how it works and how it might be useful in a number of ways.

Writing, running, and testing awesome Ansible content with natural language and AI - powered by Ansible’s MCP server

Speaker: Shatakshi Mishra

Slides: Presentation

Video: YouTube

Ansible content developers lose hours each day to context switching, which kills productivity and increases the risk of human errors.

We’ve integrated an AI-powered Model Context Protocol (MCP) server directly into the Ansible VS Code extension to address this problem. The result is a single, unified development experience that goes beyond an ordinary AI code assistant. Adding MCP server capabilities to the Ansible VS Code extension gives you an intelligent development environment that allows you to work within the context of all your existing Ansible content, including playbook, roles, and inventories. As a result, teams can reduce fragmentation in their workflows to gain productivity and standardise and accelerate Ansible content development.

Questions

  • How many tokens does it use?
    • All of the tooling is shared so the token count is not excessive. This feature is still in the process of being developed.

Reliable Network Backups & Restore (AI/ML Diff Severity)

Speaker: Rohit Thakur

Slides: Presentation

Repository: GitHub

Video: YouTube

In network automation, reliable backups are critical but often overlooked. This session explores how to design a consistent, vendor-neutral backup and restore workflow using Ansible.

Backups that “usually work” aren’t good enough. This talk presents a battle-tested, backup/restore workflow for Cisco IOS/IOS-XR/NX-OS using Ansible and Validated Content. We’ll show how to pick the right transfer method (SCP/SFTP/TFTP), enable and secure internal file servers, and make tasks idempotent across vendors. We’ll also cover selecting device-native paths (bootflash:/, disk0:, /misc/config) to avoid path pitfalls.

We will also go through a lightweight, offline rules + ML approach that scores the severity of configuration diffs (e.g., VLAN/ACL/BGP/interface changes) to prioritize reviews and automate safe rollbacks. Live demos include copying configs, verifying hashes, ranking diff severity, and controlled restores.


Check out the other sessions at CfgMgmtCamp 2026 using the links below:

Here are links to all the talks on YouTube as well as related forum discussions:

4 Likes

I really enjoyed this talk from @shamishr about the Ansible Dev Tools (ADT) MCP server and using these inside Visual Studio Code.

I think this opens up a whole new potential for Ansible content development and troubleshooting, using natural language and not needing to run a single command in the terminal :slightly_smiling_face:

Imagine this scenario, you’re developing some Ansible content and running it via Ansible Automation Platform (AAP). You have this MCP for ADT and an MCP integrated with AAP. You’re an engineer troubleshooting a job template failure in AAP, you ask your AI agent about the failure in AAP and it grabs the details via the AAP MCP. You then ask your agent to identify the failing playbook and it proposes a fix. You review and approve the fix and the ADT MCP runs ansible-lint but it identifies a lint error. The agent sees the error, proposes the fix and you approve it and commit it to your SCM which is synced as a project in AAP. You then tell your AI agent to re-run the failed job in AAP and the issue is resolved.

Thinking about it more, if you had a ticketing system like ServiceNow which created an incident for the job template failure in AAP, you could even have an MCP for that and record the fixes in the work log and automatically resolve the incident once the job succeeds again in AAP.

1 Like