How to use the Diffusion MCP Server

In the fast-moving world of real-time data, bridging the gap between streaming platforms and AI agents is no longer a futuristic dream, it’s happening now. Enter the Diffusion MCP Server from DiffusionData, a powerful new layer that transforms how you interact with streaming data. Rather than simply navigating complex APIs or sifting through documentation, you can connect your AI assistant, like Claude or AI Agents – directly to your real-time data infrastructure. 

The Model Context Protocol (MCP) is an open standard that lets AI assistants interact with external systems through a standardised interface. Rather than just discussing systems, they can actually query them, manipulate them, and work with them in real-time. 

Once connected, the MCP Server opens up two complementary ways to use AI with Diffusion … Watch the short demo below. 

Understanding the Diffusion MCP Server in Depth

Our videos below will guide you step-by-step through setting up the MCP Server, linking it to the Diffusion platform, and exploring two major use-cases: helping developers learn and build faster, and supporting operations teams with conversational monitoring and diagnostics. Whether you’re building a streaming analytics engine, a market-data feed, or simply want to unlock AI-driven control of your real-time data layer, this guide will help you get there.

Follow the links below and use the example prompts to discover the power of the Diffusion MCP Server:

1. Getting started with the Diffusion MCP Server:

Watch video here

Prompt: “Connect to Diffusion” – Connects to Diffusion using a specific URL

Prompt: “Connect to Diffusion using the URL ws://localhost:8080/” – Connect to Diffusion using the URL ws://localhost:8080/

Prompt: “Connect to Diffusion using the URL ws://localhost:8080/ and the credentials admin and password” – Connects to Diffusion using a specific URL and credentials

All commands support natural language variations. For example: – “Connect to Diffusion” can also be phrased as “Please connect to my Diffusion server” – “Create a topic at…” can be “I need a new topic at…”

2. Using Diffusion MCP Server for Topic Views:

Watch video here

List and Create

Prompt: “Can you list the current topic views” – Returns a list of the current topics in Diffusion and their output destination paths

Prompt: “Can you create a topic view with the name {topic-view-name} which mirrors the topic at the path {source/path} with the output as {destination/path}” – Creates a topic view displaying the same value of a topic at a different path

Update and Delete

Prompt: “Can you update the value at {source/path} and test the value of the {destination/path} is updated” – Tests that topic view mirroring is working correctly

Prompt: “Can you delete the topic view named {topic-view-name}” – Deletes a specific topic view

Prompt: “Can you delete all topic views” – Deletes all topic views

Type Conversion

Prompt: “Can you create a topic view with the name {topic-name} which converts {source/path} to a type of JSON” – Creates a new topic view with a specified name, which converts the type of a specified topic to JSON

Prompt: “Can you remove the topic view {view-name} and check the corresponding topic destination is removed from the topics list” – Removes a specified topic view and checks the destination topic is also removed from the list of topics

Prompt: “Can you create the source topics needed to create a topic view that converts each type” – Creates the topics need to demonstrate topic type conversions

Prompt: “Can you create topics views to demonstrate conversions between each of these topic types” – Creates topics views which will demonstrate topic conversions to all supported types using topics created in the previous command

Path Separators and Properties

Prompt: “Can you create the topics and topic views needed to demonstrate override properties” – Creates topics and topic views to demonstrate override properties

Prompt: “Can you create topics and topic views to demonstrate replacing path separators” – Creates topics and topic views to demonstrate path separator replacement

Filter Demonstrations

Prompt: “Can you create the topics needed and topic views to show how filter values works” – Creates a demonstration of how filter values work in topic views

Prompt: “Can you create the topics needed and topic views to show how throttled updates” – Creates a demonstration of how throttled updates work in topic views, with different levels of throttles applied

Prompt: “Can you create the topics needed and topic views to show how delayed updates work” – Demonstration of how delayed updates work in topic views, for scenarios like delayed market data for free users or creating time-shifted data streams

Insert Directive

Prompt: “Can you create the topics needed and topic views to show how insert data works” – Create a comprehensive demonstration of how the insert directive works in topic views. The insert directive allows you to pull data from other topics and insert it into your reference topics, which is great for data enrichment

JSON Patch Operations

Prompt: “Can you create the topics needed and topic views to show how apply JSON patch works” – Creates a comprehensive demonstration of how JSON Patch operations work in topic views. Shows how JSON Patch (RFC 6902) allows you to perform sophisticated JSON transformations using standard patch operations

Remove Operation

Prompt: “Can you create the topics needed and topic views to show how process data remove works” – Create a comprehensive demonstration of how the remove operation works in the process directive. The remove operation allows you to delete fields from the topic value as it’s transformed by the topic view

Set Operation

Prompt: “Can you create the topics needed and topic views to show how process data set works” – Create a comprehensive demonstration of how the set operation works in the process directive. The set operation is incredibly versatile – it can add fields, modify values, restructure data, perform calculations

3. Using the Diffusion MCP Server to understand Session Trees:

Watch video here

Prompt: “What are session trees used for in Diffusion?”

Prompt: “Can you set up topics to represent a simple market data application?”

Prompt: “Can you show me how to use session trees to deliver different prices on the same topic to different users. And can you set up a test to try it out”

Prompt: “Show me all currently connected sessions and their key details.”

Prompt: “Create a session tree that personalizes data for each client session based on their attributes or roles.”

Prompt: “Configure session trees so that each session only sees a subset of the topic tree relevant to that user.”

Prompt: “List all existing session trees and explain what each one does.”

Prompt: “Update the session tree rules to restrict access for a specific group of sessions.”

Prompt: “Remove the session tree configuration that was created in this session.”

4. Understanding Diffusion’s security:

Watch video here

5. Exploring metrics in the Diffusion MCP Server:

Watch video here

List Available Metrics

Prompt: “Can you return a list of available metrics which can be used to create alerts” – Lists all available Diffusion metrics that can be monitored with alerts

6. Exploring metric alerts in the Diffusion MCP Server:

Watch video here

Create Metric Alerts

Prompt: “Create a metric alert for os_system_cpu_load at the path ‘OS/CPU-Load'” – Creates an alert to monitor operating system CPU load

Prompt: “Can you create a metric alert that creates an alert if the select diffusion_topics_current metric is above 5 please” – Creates an alert that triggers when current topic count exceeds 5

Prompt: “Can you create 6 topics to trigger the above alert and check the alert is created” – Creates test topics to verify the metric alert triggers correctly

Prompt: “Can create a metric alert for jvm_threads_current invoked when the value is above 200 with the alert at alerts/threads/high_count” – Creates an alert for high JVM thread count (>200 threads)

Prompt: “Can you create a metric alert for diffusion_topics_subscribers into topic alerts/topics/high_subscribers where value > 500” – Creates an alert when topic subscriber count exceeds 500

Prompt: “Can you create a metric alert for diffusion_server_free_memory_bytes into topic alerts/memory/low_free where value < 100000000” – Creates an alert for low free memory (<100MB)

Prompt: “Can you create a metric alert if jvm_threads_deadlocked is above 0 with the alert topic at alerts/threads/deadlocked” – Creates an alert to detect JVM thread deadlocks

Prompt: “Can you create a metric alert if diffusion_topics_subscribers is higher than 500 with the alert topic as alerts/topics/high_subscribers” – Creates an alert for high subscriber count (alternative phrasing)

Prompt: “Can you create a metric alert where os_process_cpu_load goes above the value 0.01 at the topic path alerts/cpu/higher-process” – Creates an alert for elevated process CPU usage

Manage Metric Alerts

Prompt: “Can you change the value of the os_process_cpu_load metric alert to the value of 0.005” – Updates the threshold value for an existing metric alert

Prompt: “Can you remove the metric alert for {name-of-metric}” – Removes a specific metric alert by name

Prompt: “Can you remove all metric alerts” – Removes all configured metric alerts

7. How the Diffusion MCP Server creates topics to represent a simple market data application:

Watch video here

Prompt: “Create topics that represent a simple market data application, including FX, stocks, and indices under a ‘market’ hierarchy, using JSON topics with realistic example data.”

Prompt: “Create a time series topic for historical data, suitable for maintaining a history of price updates, and populate it with some example historical events.”

Prompt: “Create a time series topic that keeps the last 100 events, with new subscribers receiving the last 10 events on subscription.”

Prompt: “Query the historical time series topics to show the stored data in a clear, readable format.”

Prompt: “Create a simple topic view over the market data that aggregates all stock prices into a single portfolio summary topic showing total portfolio value and performance.”

Prompt: “Create another topic view, a simple market overview that flattens the hierarchy for easier consumption.”

Prompt: “Create topic views for a market dashboard, asset-by-type view, and flat market view, without duplicating underlying data.”

Prompt: “Remove all the topic views that were created in this session.”

Prompt: “Remove all the topics that were created in this session and then disconnect from the Diffusion server.”

If you want to find out more about how the Diffusion MCP Server works, you can get in touch with us here.