5 posts tagged with "MCP"

EventCatalog is good at telling you how your architecture works today. Your teams document their domains, services and messages, and everyone can find out what talks to what.

But the conversations that matter most are often about tomorrow. "How would we add buy now, pay later?" "What happens if we split the Order Service?" "Who needs to know if we retire this API?"

Those conversations usually happen somewhere else: a whiteboard, a slide, a diagram tool that knows nothing about your architecture. Someone redraws the services by hand, gets a few of them wrong, and the result ends up in a folder nobody opens again.

We kept hearing the same thing from customers: we document the present in EventCatalog, but we plan the future somewhere else.

Today we're excited to introduce EventCatalog Studio (v2): live, collaborative canvases inside your catalog, for designing what comes next with your team and your AI agents.

The first version of EventCatalog Studio was a separate tool for drawing designs with architecture primitives. Studio v2 lives inside EventCatalog itself, so your designs start from the architecture you already document, and your team and your agents can work on them together.

Ask an AI tool how a service works and you get a wall of text.

"The Order Service accepts PlaceOrder, publishes OrderCreated and OrderCancelled, which are consumed by Fulfilment and Notifications, which..." By the third service, you are drawing the diagram in your head.

Some tools will draw one for you. But a diagram an AI model generates is a guess. It can look right and still be wrong, and you have no way to tell without checking it against the real thing.

You already have the real thing. Your catalog knows exactly how your services, messages and domains connect.

In EventCatalog v5, the MCP server can show your actual architecture diagrams and schemas inside the conversation. In Claude, ChatGPT, VS Code and other tools that support MCP Apps, you get the interactive EventCatalog visualiser right in the chat.

More context is not always better context.

EventCatalog's MCP server already lets AI tools and agents query your entire architecture. That is useful when you need to understand organization-wide dependencies, but most engineers spend their day working inside a much smaller boundary.

You can now connect an MCP client to a specific domain or system. Your team gets the architecture knowledge it needs, including the connected services, messages, schemas, flows, language, and decisions, without pulling unrelated parts of the catalog into every conversation.

Your catalog already knows where your APIs live. It knows which services expose OpenAPI, AsyncAPI, or GraphQL specifications. It knows which domains describe broader API surfaces.

EventCatalog can now turn that knowledge into a standard discovery endpoint at /.well-known/api-catalog.

I'm excited to share that we've improved and simplified our MCP server for EventCatalog.

Previously, to use EventCatalogs MCP server you had to install and run the standalone @eventcatalog/mcp-server package separately, the previous MCP server would use the llms.txt file to provide context to the LLM. Although this worked, it was not the most efficient way to use EventCatalog with MCP clients...

We have reviewed the MCP server and made it much more efficient and easier to use.

Every EventCatalog instance now includes a built-in MCP server at /docs/mcp/ and now uses more performant ways to provide your models with the information they need.

This means you can connect your preferred MCP clients directly to your EventCatalog instance and start asking questions about your architecture in natural language.

Why use EventCatalog's MCP server​

EventCatalog is documentation for both humans and AI. Your teams browse the catalog to understand services, events, and domains. Your AI tools can now do the same.

The MCP server gives LLMs and MCP clients (Claude Desktop, Cursor, Windsurf, VS Code) direct access to your architecture documentation. Ask questions about message schemas, trace service dependencies, or analyze change impact.

The AI works with the same source of truth your team uses.

This is the best of both worlds. Humans get a visual catalog they can browse and search. AI gets structured access to query, filter, and reason about your architecture. Both stay in sync because they read from the same place.

Getting started​

To get started, you need to update your EventCatalog instance to the latest version and configure it to run in SSR mode.

First you need to update your EventCatalog instance to the latest version.

npm install @eventcatalog/eventcatalog@latest

Then you need to update your eventcatalog.config.js file to use SSR mode. (mcp requires SSR mode)

export default {
  output: 'server',
  // ... other config options
}

Once EventCatalog is running you can connect your preferred MCP client to your EventCatalog instance at /docs/mcp/.

To get started, see our MCP server documentation.

Example usage​

Once connected, you can ask questions about your architecture in natural language. This can be useful for:

• Finding messages for features you're building
• Accessing catalog directly in code editors
• Defining standards and creating schemas
• Exploring OpenAPI and AsyncAPI files
• Understanding service impact from changes
• Discovering resource ownership

Here are some examples:

"What's the blast radius if I change the OrderCreated event?"

The OrderCreated event (v2.0.0) is consumed by 4 services:

Service	Team	Impact
InventoryService	Platform	Updates stock levels
NotificationService	Communications	Sends order confirmation
AnalyticsService	Data	Tracks order metrics
BillingService	Finance	Initiates payment flow

Total blast radius: 4 services across 4 teams. Consider coordinating with Platform and Finance teams before making breaking changes.

"Which services does the Payment team own?"

The Payment team owns 3 services:

• PaymentGateway (v1.2.0) - Processes credit card transactions
• RefundService (v1.0.0) - Handles refund requests and chargebacks
• PaymentWebhooks (v2.1.0) - Receives payment provider callbacks

These services produce 5 events and consume 3 events total.

"Show me the schema for UserRegistered"

Here's the UserRegistered event schema (v1.0.0):

{
  "userId": "string (UUID)",
  "email": "string (email format)",
  "registeredAt": "string (ISO 8601 datetime)",
  "source": "enum: web | mobile | api",
  "marketingConsent": "boolean"
}

This event is produced by AuthService and consumed by NotificationService, CRMSync, and AnalyticsService.

"Explain the checkout business flow"

The Checkout flow consists of 5 steps:

1. Cart Validation - CartService validates items and availability
2. Price Calculation - PricingService applies discounts and taxes
3. Payment Processing - PaymentGateway charges the customer
4. Order Creation - OrderService creates the order record
5. Fulfillment Trigger - Emits OrderCreated to start shipping

Average completion time: 2.3 seconds.

Tools and resources​

The MCP server includes 15 tools for querying your architecture and 9 resources for bulk data access.

View all tools and resources →

Migrating from standalone server​

The standalone @eventcatalog/mcp-server package still works, but we plan to deprecate it in a future release. The built-in server is faster, uses less context, and requires no external setup. We recommend migrating to the built-in MCP server when possible.

Summary​

The built-in MCP server simplifies deployment by eliminating external server management. It provides faster queries through direct content collection access and supports custom tools for domain-specific integrations.

The built-in MCP server requires SSR mode and a Scale license. Get a free trial at eventcatalog.cloud.

Learn more:

• MCP server documentation
• Complete API reference

Get help:

• Discord community
• GitHub issues