Macktez
Our WorkCase Studies

Technology & Software

Macktez Monitor: Building on Our AI Foundations

How we turned the connections behind our AI tools into a monitoring platform — and brought its history and insights back to those tools.

Macktez (internal)AIIdentitySecurityCloud & Infrastructure
The Macktez Monitor systems diagram. In the upper left, the vendor platforms we read, each connected to the shared gateway at mcp.macktez.com at the center, with the access control beside it. On the right, the collection path: the roster and its safeguards, the fleet collector, and the report store on Amazon S3, with the two collectors inside a client network feeding the same store. Below, the Monitor application: readings received and normalized, judged by deterministic rules, written to a Postgres database, and read back by Team Members through the website and by Claude and support.agent through the Monitor interface.

At a Glance

Client
Macktez (internal)
Problem
No single dashboard showed everything we wanted to know. Vendor consoles each exposed part of the picture, with limited history. Understanding a client’s environment meant gathering information from several systems by hand.
Approach
Reuse our existing MCP connections to collect data on a recurring schedule, retain its history, and combine it into a more complete picture.
Outcome
A growing monitoring platform that gives our team and AI tools shared access to system health, historical context, and insights that individual platforms cannot provide on their own.

The Idea

Existing infrastructure, new capabilities

We built our custom Model Context Protocol (MCP) connections so our team could use AI tools to work across the platforms we manage. That shared foundation subsequently enabled support.agent, our autonomous digital employee.

It also created an opportunity. The connections could serve ordinary software as readily as a language model. We could use them to gather information continuously, without building another set of vendor integrations.

Using AI-assisted development, we built Macktez Monitor around that idea. A scheduled script queries our vendor platforms through MCP, while collectors inside client networks gather readings from local systems. Reports arrive in a central storage repository on AWS S3, where they trigger processing by the Monitor application.

Deterministic rules interpret the readings and store the results. No LLM is involved in collection or health assessment.

The Macktez Monitor systems diagram. In the upper left, the vendor platforms we read, each connected to the shared gateway at mcp.macktez.com at the center, with the access control beside it. On the right, the collection path: the roster and its safeguards, the fleet collector, and the report store on Amazon S3, with the two collectors inside a client network feeding the same store. Below, the Monitor application: readings received and normalized, judged by deterministic rules, written to a Postgres database, and read back by Team Members through the website and by Claude and support.agent through the Monitor interface. Along the bottom, the rules for network evidence, backups, servers and directories, and the six states.
Every box is a system that exists and every arrow is a real path between two of them, numbered in the order a reading travels: from the vendor platforms through the shared gateway to the collector, into the report store, through the rules into the database, and out to the people and assistants who read it.

Open the full-size version to read the individual panels.

What It Adds

A more complete view across systems and time

Bringing the data together makes new capabilities possible.

A live network topology map, for example, draws on information from firewalls, switches, access points, and device management platforms. Assembling that view previously required logging into several systems and reconciling their answers. Monitor now renders it automatically, showing connections across equipment from different vendors and making gaps in the available information visible.

Monitor’s topology view for one office. The internet sits at the top, two carrier circuits are marked active and standby, a pair of firewalls feeds a switch stack, and the wireless access points hang below it. A side panel shows the switch’s ports as a grid colored by VLAN and outlined by link speed, with one port selected and the device on it named.
One site, mapped automatically by combining readings from firewalls, switches, access points, and network probes. The panel on the right shows individual switch ports and their connections.

History adds another dimension. We can see when a device disappeared, whether a condition is recurring, or how long it has been since a backup completed successfully. A successful reading also has an age: when fresh information stops arriving, Monitor shows that uncertainty instead of leaving yesterday’s reassurance on screen.

Together, those capabilities give our team a stronger starting point for troubleshooting and planning.

Closing the Loop

Better information feeds back into our AI tools

Monitor exposes its accumulated history and insights through its own MCP tools. Our team, support.agent, and our diagnostic workflows can use that context alongside live vendor readings.

An investigation can begin with what changed, what has persisted, and what else was happening at the site, then query individual systems for more detail.

That closes the loop: infrastructure built for AI enables a new operational tool, and that tool gives AI-assisted work better information.

Built to Last

Built to improve and expand

Ongoing performance work keeps Monitor lean as its coverage grows. We measure changes, reduce unnecessary processing and database writes, and refine how much detail we retain over time.

That discipline keeps operating costs low and creates room for more systems and more frequent readings. Each addition to our shared connectivity becomes another opportunity to extend the platform.

AI-assisted coding helps us act on those opportunities: developing features, improving existing workflows, and testing whether an idea delivers enough practical value to keep.

For Clients

Find new solutions to familiar challenges

Monitor began with a familiar frustration and infrastructure we already had. Recognizing how those pieces could work together created something more useful than either alone.

We help clients find similar opportunities in their own businesses: understanding where work gets stuck, identifying where AI and custom software can help, and building solutions that fit the way their teams operate.

Start With What You Have

Ready to explore what we could build together?

Tell us where the work gets stuck. We’ll help you see where AI and custom software fit the way your team already operates.

Let’s Talk

Related

For the connections underneath it and the network work it watches, see:

MCP ToolingView ServiceNetwork ManagementView Service

For the fundamentals behind what it measures and how its access is scoped, see our University guides:

Network FundamentalsRead ArticleThe Principle of Least PrivilegeRead Article

For the rest of what we built on these connections, and how we build, see these case studies:

Meet support.agent: Our First Fully Autonomous Digital EmployeeRead Case StudyOperationalizing AI Across Every DepartmentRead Case StudyVibe Coding, Done RightRead Case Study