Demos

Network Visualization Through Natural Language Using FlowAI – No Parsing, No Templates, No Code

Video Notes

(So you can skip ahead, if you want.)

0:00 MCP Installation & Configuration
0:49 FlowAgent Build with Natural Language Prompts
1:44 Agent Execution & Results
2:29 Executive Overview Mind Map
3:21 Interface & Addressing Mind Map
3:48 Routing Table Mind Map & On-Demand Refresh

Timely Data Visualization Using Natural Language with FlowAI

By leveraging the power of the markmap.js.org Model Context Protocol (MCP) authored by John Capobianco, our FlowAI Agents have unlocked a streamlined approach to data visualization. This capability allows agents to effortlessly convert information into interactive mind maps using nothing more than natural language instructions.

In this demo, John builds a FlowAgent that discovers network infrastructure with PyATS Genie and generates three comprehensive visualizations in under 2 minutes. A simple prompt – “Perform a full network discovery and generate visual documentation” – triggers the agent to interpret intent, gather data, and produce structured mind maps without any intermediate coding steps.

What You’ll See:

First Node.js MCP integration running in production on the IAG platform
Custom FlowAgent combining PyATS discovery tools with MarkMap visualization
12 automated tool calls delivering executive overviews, interface maps, and routing documentation
Real-time refresh capability – update network state on demand with zero manual work

Why It Matters

Traditional network documentation requires parsing show commands, writing regex patterns, building Jinja2 templates, and maintaining fragile automation scripts. FlowAgents eliminates all of it.

Now, a simple request like “gather the state of my device interfaces and transform them into a mind map” is all it takes. The agent directly interprets your intent and generates beautiful, structured visualizations – making complex data instantly understandable and accessible. Build the agent once, run it anytime, get documentation that’s always current because it’s generated fresh from live network state.

–+

John Capobianco • 00:00

I’ve just made a new agent in FlowAgents called Network Visualization as Mind Map. Alright, so we’re gonna look at the agent, but 1st I wanna show you that on the IAG platform, I did a get clone and brought in the Mark Map. Customized for this platform MCP. I registered it here, and I can run it here and test it here. It is the 1st Node.js MCP that I’ve installed, and it runs just fine. So I’m going to actually switch into the MarkMap output folder, and we’re going to switch away from the terminal because I actually have it connected here in VS Code. So we’re going to keep an eye on this MarkMap output folder.

John Capobianco • 00:49

Excuse me. And here’s the prompt. So the system prompt that you’re, you know, an IE-level network automation agent specializing in infrastructure discovery and documentation using PyATS Genie, its prime responsibilities, what I want it to gather, and then the actual user, right? Perform a full network discovery and generate visual documentation for the device, the device name, critical instructions to use the MarkMap output tool just to make it a little more deterministic to call the MarkMap workflow. We’ve bound the PyATS show command and we’ve bound the project. And in the project, I have this very deterministic workflow with the variables needed to connect this MCP for visualization. So let’s go ahead and run the agent.

John Capobianco • 01:44

And I am going to just come back because we don’t need to sit here and watch it churn, but it does only take a few minutes. So I’ll be right back. Okay, so the flow agent is done and it took about 2 min , 12 tool calls, and some of these are the PyATS show commands, as you would expect. And then further down, though, we’re calling the mark map tool. And you can see that this is for the network discovery 3. So I asked it for three different reports, and I did get three different. Here’s the summary of everything that it did, and the mark maps generated: executive overview mark map, interface and addressing mark map, routing table mark map, key findings, right?

John Capobianco • 02:29

So here are the different mark maps that we got. And let me try to make this a little easier for everyone to see. Right, so this is the platform information, the interface summary, and some key interfaces about this device. Right, in less than 2 min , it discovered all of this information with PyATS and used the mark map. Now, we’re going to go to one of the more detailed mark maps here, and we can just collapse that for the sub-interfaces. And we have all the sub-interfaces with their IP address and status here. We’ve got loopback interfaces, physical interfaces, virtual interfaces, unassigned sub-interfaces, some sub-interfaces without IP addresses.

John Capobianco • 03:21

So, it’s trying to give us some insight, even. It’s trying to help us. And then the routing mark map. And you saw me build the agent and run the agent. So the point-to-point links, everything on connected routes, everything in local routes, the dynamic routes, static routes, and a routing summary. Right? Pretty cool.

John Capobianco • 03:48

Just from the agent. Now I can refresh that at any time by just managing the agent. and running the network topology, running the network visualization agent. I just hit run agent and it will refresh that. Again, the tools I’m using, a couple of PyATS tools, and my project, which has this deterministic workflow component to make the folder and where to store the data and how to connect it back to the IAG platform. All right, thanks again.