Gartner® Recognizes MetricsHub® for Robust and Scalable Metric Collection. Read more

Blog

Meet M8B: A Virtual Assistant for Sys Admin Teams

Ever wished for a grumpy but brilliant sysadmin intern? We built one. Discover how M8B, powered by MetricsHub® and AI, helps IT teams tackle issues fast.

More Time to be an Architect, Less Time as a Firefighter

System administration involves countless routine checks, verifications, and repetitive tasks. While essential, they can quickly become a burden in the daily grind. And let’s be honest: every sysadmin has, at some point, wished for a personal assistant (or at least an intern) to handle the boring stuff while focusing on the interesting, high-value work.

At MetricsHub®, we worked to make that dream a reality by creating a virtual assistant with the expertise of a senior admin. That’s how M8B (maybe) was born. Inspired by our love of Star Wars, its name follows the droid naming style while nodding to R2-D2.

In the following sections, we’ll share how we built M8B with MetricsHub, the technologies that power it, and real-world stories of how it supports human sysadmins.

The Technical Stack

To bring our Virtual Sys Admin to life, we combined six key technologies, each playing a distinct role in the ecosystem:

  • MetricsHub and its MCP server: The watchful eye and hand on the IT infrastructure
  • ngrok: The secure gateway to the outside world
  • Zapier: The behind-the-scenes conductor
  • Vector Store: The institutional memory
  • ChatGPT: The conversational genius
  • Slack: Where the virtual sys admin meets the team

alt text

MetricsHub and its MCP Server: The Watchful Eye on the IT Infrastructure

The Model Context Protocol (MCP) is an open standard that allows AI models to securely connect with external tools and data sources. By integrating an MCP server into MetricsHub, we gave any LLM the ability to:

  • Request real-time metrics from MetricsHub
  • Perform any manual check on the infrastructure
  • Gather detailed information about any server, network device, or storage system.

With MCP, your everyday chatbot gains both the eyes and hands of a sysadmin.

ngrok: The Secure Gateway to the Outside World

We relied on ngrok to create a secure tunnel from local environment to the internet, making internal services accessible without exposing them directly.

And the best part? We didn’t have to touch our firewall or network settings. With this setup, any LLM online can communicate with MetricsHub.

Zapier: The Behind-the-Scenes Conductor

Zapier orchestrates the flow of information between Slack, ChatGPT, and MetricsHub, ensuring every step runs smoothly.

  1. The workflow starts when a message is posted in the IT Slack channel
  2. Zapier extracts the Thread ID to ensure all replies in the same thread are tracked together.
  3. Zapier retrieves the entire conversation from that thread
  4. Zapier summarizes the collected messages in a clear prompt for ChatGPT to make sure the LLM has the full story
  5. ChatGPT generates a helpful response. This is where M8B, our Virtual Sys Admin comes into play
  6. Zapier converts the output into a Slack-compatible formatting.
  7. Finally, Zapier posts the response back into the IT Slack Channel.

Zapier keeps the prompt delivered to ChatGPT clean, making the conversation feel as natural as possible. While it already works well, there’s room to grow, for example in handling interruptions or edited messages.

Vector Store: The Institutional Memory

To make M8B even more reliable, we added a vector store containing all our current documentation about IT architecture and internal services. This allows ChatGPT to go beyond raw metrics and ground its answers in verified knowledge from our internal docs.

Instead of guessing, M8B can retrieve the right context in real time and turn them into precise, relevant responses.

ChatGPT: The Conversational Genius

M8B is based on GPT-5, and we could have relied on any LLM. Once connected to MetricsHub, any model can become a powerful assistant for system administrators. In our stack, ChatGPT’s role is to understand requests, suggest fixes, and explain what’s happening in plain language. Paired with MetricsHub, it turns raw metrics into clear, actionable responses.

To ensure reliability, we defined strict configuration rules:

  • Answers must rely only on verified data from the MetricsHub MCP server, our IT knowledge base, or details provided by the user.
  • If uncertain, M8B says “I don’t know or labels its reasoning with “Guess:”.
  • It cannot promise future actions or simulate operations it cannot perform.
  • It replies in the user’s language (English or French).
  • Explore our conversation rules

And because no sysadmin is complete without a bit of personality, we made M8B a grumpy but highly competent system administrator. This touch of realism made interactions not just useful but also fun.

Slack: Where the Virtual Admin Meets the Team

Slack is the interface where employees report problems, ask questions, and receive guidance. By integrating MetricsHub and ChatGPT with Slack, we turned the familiar IT incident channel into the front desk of IT Team’s operations.

No new tools to learn, no extra logins to manage: M8B simply blends into existing team conversations, ready to jump in whenever needed.

Hearing Stories from the Field

A tech stack is only as good as the results it delivers. M8B proved its worth the very first time we put it to the test in a live environment.

During a team discussion, M8B spotted and resolved two critical server issues in just a few minutes. The intervention was so quick that one of our sys admins was surprised at first, but quickly convinced. But the results spoke for themselves: downtime avoided, users unblocked, and a lighter load for the human team.

Here’s how it unfolded:

alt text

  1. LLM tests were running unusually slow, and a user wondered aloud whether our Nvidia server was functioning properly.
  2. M8B immediately flagged an issue on the server, provided a detailed explanation of the problem, and suggested actions to fix it
  3. Still skeptical, the user called in his human colleague Pascal for confirmation.
  4. Pascal joined the thread, reviewed M8B’s findings, and after a few exchanges with teammates:
    • Checked that the libraries matched the loaded driver
    • Ran a few diagnostic commands
    • Got the LLM tests running smoothly again
  5. M8B closed the loop by verifying that everything was back to normal and confirming that Nvidia server was responding correctly.

What started with doubt ended with trust: M8B had not only detected the issue but also guided the resolution step by step, saving time and frustration for everyone involved.

Conclusion

M8B has shown that a virtual sys admin makes IT teams faster, more resilient, and far less burdened by routine tasks. It doesn’t replace humans, it empowers them with instant insights, quicker troubleshooting, and fewer disruptions. The payoff: less downtime, smoother operations, and IT teams free to focus on what really matters.

Configure AI Agent Integration Create Your Virtual Assistant

Share this post