Hi, everyone, my name is Mark McBride and I'm a customer success specialist. And today we're looking at the model context protocol or MCP server in Nexus dashboard.00:07MCP lets AI clients and large language models work directly with your fabric telemetry. And the goal is to get answers faster without having to spend much time writing code. Today we'll cover the following four topics, starting with the architecture and why it matters operationally.00:17Then we'll take a look at the types of tools and telemetry the MCP server exposes, followed by how to connect an AI client to Nexus dashboard. And then we'll jump into a live demo to see how it all works. Let's 1st start with the why.00:36Using MCP allows AI clients to connect to Nexus dashboard using a standardized integration. So you don't have to write REST calls, handle authentication, and parse through JSON files for each case. The MCP server exposes ready to use tools and clients like cursor, VS code, and other agents.00:53Make use of those tools when interacting with Nexus dashboard.01:13Using MCP also provides operational efficiency. You can ask plain English questions like which fabrics have critical anomalies right now, and the model picks the right tools and then pulls together symptoms, root cause graphs, and the fabric topology.01:17And security. Every call goes through the Nexus dashboard API gateway, so our back, tendency, isolation, and triple AAA all still apply. The AI agent can only see what your user account can see. Next, we'll talk about the boundaries. So Nexus dashboard MCP.01:33Is read only by design or every tool can only collect data. There are NO configuration changes and NO policies that can be pushed to devices.01:53Support for Nexus dashboard MCP is available starting in version four.three.one. It is currently supported for land fabric deployments and Nexus dashboard Federation for multi cluster is currently not supported in this version.02:02And the telemetry data comes from Nexus dashboard, but the LLM is what does the interpretation. So operators still remain accountable for verifying the output against the data in their Nexus dashboard. And here's the full tool set grouped by purpose.02:17Starting with health and inventory, you can use the tool, get ND fabric summary, which will give you overall health of all the fabrics that have been discovered. Or you can use the get fabric details, which will drill into inventory, interfaces, and endpoint summaries.02:36There's also anomalies and root cause. The anomalies, summary, groups, events, by severity and category defaulting to the last 2 h. Or you can get anomaly details, which will return the root cause impacted objects in the node graph.02:53When it comes to AI and machine learning telemetry, these tools track GPU allocations, job runtime, and time series performance, which is essential if you're running an AI backend fabric. When it comes to assurance, network analytics surfaces things like congestion and PFC drops, which are critical for.03:10The Lossless Rocky environments.03:30And topology and segmentation. Here you get the unified topology graph and things like tenants, vrfs, EPGs, contracts, and L three outs. Onboarding can be done in four steps, starting by generating an API key.03:33By going to the admin and manage API keys in Nexus dashboard and make sure to copy it right away because it's the only time it'll be shown and RBAC applies per user, so pick an account with the right scope. Then you'll want to configure your mCP.JSON file within your AI client such as cursor VS code.03:50We'll point it to the URL of your Nexus dashboard and you'll use the base URL of/API/V1/MCP. And also make sure to add the API key and the username and the headers.04:11For production environments, you'll want to make sure to install the CA certificates on both Nexus dashboard and your client's trust store. And when it comes to pagination, it's handled for you. The max parameter ranges from one to 10000 with a default of 100.04:24So out of range values are corrected automatically. Alright, let's now move into the demo where I'll walk you through what we've just discussed. Let's take a look. Let's 1st start off by logging into Nexus dashboard.04:40And in this instance, we're currently running Nexus dashboard four.three.one, which includes support for using MCP. Once we're logged in, we're gonna want to create an API key.04:57That we can use for our MCP connection from cursor to Nexus dashboard.05:17So we'll go to admin and then we'll go to manage API keys, then we're gonna add an API key. We'll give it a description, then we'll select save.05:23And then we'll want to copy off this key. This is the only time we'll be able to copy this key and then we'll treat it just like a password and make sure you copy this key off someplace secure.05:39Now that we've got our key, we'll then head over to cursor to set up our MCP connection.06:04In cursor, we're gonna use an existing repository we created for our project here, and then we're gonna add an agent, and then if we select the plus sign and then go to MCP.06:20We can currently see that there are NO MCP servers set up for this repository, so we can go to manage.06:32And then we can select the option to add a new MCP server, and then we're gonna add an MCP server to our workspace Nexus dashboard MCP. So now this creates the mCP.JSON file that we use to connect to Nexus dashboard.06:39This.06:59So in our MCP.json, we're gonna fill in the following information. We're gonna give it the IP address of our Nexus dashboard server.07:05Yeah, for the API key, we're gonna paste in the API key that we created from Nexus dashboard.07:21Yeah for the ND username, this will be the user that created the API key.07:37And then we'll save our MCP.json file Now that we have our MCP.json file created, we can now see that we have an option for the.07:52The Nexus dashboard MCP server available in our workspace.08:12From here we can now enable our connection and here we can see we are now connected to the Nexus dashboard MCP server, and here.08:27You can see all of the available tools that are provided from the Nexus dashboard MCP server. So whenever we ask questions like.08:42Give me, a summary of the Nexus Dashboard fabrics or show me a list of anomalies. These are the list of tools that are available for cursor or your AI client to use with Nexus dashboard.08:51So now that we're back in agent mode, when we click on the plus sign and go to MCP, we can see we have our connection that's enabled for our ND MCP server. So what we can do from here is now.09:13Ask questions about our Nexus dashboard environment. E.g., we can ask to provide a list of all fabrics from Nexus dashboard.09:32And what the agent will now do is use the tools that are available to connect to Nexus Dashboard and then pull a summary of all of the fabrics that it's discovered from Nexus dashboard.09:47Here we can see the tool that is being used to list Nexus dashboard fabrics and their health summary. So here we can see that the agent was able to connect to Nexus dashboard and pull the two fabrics that have been onboarded into Nexus dashboard. We can see what type of fabric they are. We can also see the current state of a.10:08Anomalies, if there are any advisories within the fabric, as well as their config sync status, their software versions and their BGP information.10:28So let's get a few more details about these fabrics. So let's ask for a list of switches and what software version they're currently running.10:45The agent will now go and find the available tools to be able to pull these details from the switches and then connect to Nexus Dashboard using those tools to give us a list.11:14List of all the switches and the software versions that are in both fabrics.11:25So here we get a breakdown of each fabric, the switches that are in each fabric, what roles they're running in and the software version for for each switch in each fabric.11:33Next let's asks questions about advisories and anomalies. So let's say create a PDF report of all advisories and anomalies.11:54For all fabrics and the AI agent will now go and use the tools that are available for collecting anomalies and advisories and then create a PDF report based on what it's found.12:09And.12:28And here we can see that there are five different tools that the agent is going through to create our report.12:49The agent has now finished collecting all of the anomalies and advisories and has created a PDF report based on its findings. In this report, we can see the advisories and anomalies for each fabric.13:01We can see the total number of anomalies, one critical for our classic land fabric, NO active anomalies for our VXLAN EVPN, and we can see the number of advisories for each fabric. The anomalies we can get a detailed list of each anomaly that was detected.13:21As well as the advisories that had found that can be used for best practice and hardening recommendations for our Nexus dashboard fabrics. So in this demo, we saw how we can use an AI agent such as cursor as well as Nexus dashboard for.13:40Two.three.one that has the built in MCP server. That way we can use AI agents to be able to interact with our Nexus dashboard fabrics. Thanks for watching.14:00
