Anthropic and OpenAI officially launched MCP Apps, the first extension to the Model Context Protocol. This is a big deal. For the first time, AI tools can return interactive UIs, not just text, directly in your conversation with Claude, ChatGPT, or any MCP-compatible client.
And if you're wondering how to build, test, and debug these new capabilities? Postman already has you covered.
What Are MCP Apps?
MCP (Model Context Protocol) standardizes how AI models connect to external tools and data. Until now, that meant text in, text out. You'd ask Claude to query your database, and it would return rows of data. Useful, but limited.
MCP Apps changes that. Now tools can return:
Interactive dashboards where you filter, sort, and drill down into data
Configuration wizards with dependent fields and real-time validation
Document viewers where you can annotate PDFs inline
Real-time monitoring with live updates
Instead of typing "Show me just the ones from last week" and "Sort by revenue" and "What's the detail on row 47?", you just... click. The AI stays in the loop, seeing your interactions and responding accordingly.
Why This Matters for API Developers
Every MCP tool is, at its core, an API integration. When Claude calls a tool to check your calendar or query your database, it's making API calls through a standardized interface.
This is exactly what Postman has been doing for over a decade: helping developers design, test, and debug API integrations. The difference now? The client isn't a human developer clicking "Send." It's an AI agent making decisions.
Postman's MCP Support (Already Shipped)
Postman has been building MCP capabilities since mid-2025, and the tooling is already robust:
MCP Client in Postman
You can create MCP requests just like HTTP, GraphQL, or gRPC requests. Click "New," select "MCP," and you're connected to any MCP server. Postman auto-discovers tools, prompts, and resources. You can test each one, inspect payloads, and debug issues, all in the interface you already know.
AI Tool Builder
This is the fastest way to go from APIs to MCP servers. Postman's network includes 100,000+ public APIs from verified sources like Salesforce, UPS, and Stripe. Select the APIs you need, click Build, and download a working MCP server. Test it in Postman, then use it in Claude, Cursor, or VS Code.
Postman Remote MCP Server
For the other direction, Postman's own MCP server lets AI agents work with your Postman workspace. Available on our Postman Github, it comes in three flavors:
Minimal: 37 essential tools for common operations
Code: Tools for searching APIs and generating client code
Full: 100+ tools covering the entire Postman API
Connect it to Claude Code, Cursor, or VS Code, and you can manage collections, run tests, and automate API workflows using natural language.
MCP Apps Support
Yes, Postman already supports the MCP Apps extension. When a server returns an interactive UI, you can view and interact with it directly in Postman's response panel. This makes Postman the ideal environment for developing and testing MCP Apps before deploying them.
A Practical Example
Let's say you're building an MCP server for your company's sales data. Without MCP Apps, Claude can query the data and summarize it. With MCP Apps, you can return an interactive chart where users:
Toggle between metrics (revenue, deals, conversion rate)
Filter by region, team, or time period
Click into specific deals for details
Export filtered data to a report
The AI sees all these interactions and can respond: "I notice you're focusing on EMEA performance. Want me to compare that against last quarter?"
To build this, you'd use Postman to:
Design the underlying API that serves the data
Create an MCP server wrapping that API (or generate one with AI Tool Builder)
Add MCP Apps metadata to return your React/Vue/vanilla JS dashboard
Test the full flow in Postman's MCP client
Deploy and connect to Claude, ChatGPT, or your IDE
The Industry Is Converging
What's remarkable about MCP Apps is the collaboration behind it. Anthropic and OpenAI worked together on the specification, building on earlier work from the MCP-UI community and OpenAI's Apps SDK. The same MCP App you build today works in Claude, ChatGPT, VS Code, Cursor, Goose (Block's agent), and more.
This is the "universal connector" vision that MCP promised. And with Postman's MCP tools, you don't need to be an AI infrastructure expert to participate. You just need to know APIs.
Getting Started
If you're already using Postman:
Update to the latest version (MCP support is in v11.79+)
Create a new MCP request and connect to any MCP server
Explore the MCP Catalog at postman.com/explore/mcp-servers
Try the AI Tool Builder to generate MCP servers from existing APIs
If you're building MCP Apps:
Check out the MCP Apps Quickstart
Use the
@modelcontextprotocol/ext-appsSDKTest your apps in Postman before deploying
The agent era is here. APIs are the foundation. And Postman is where you build and test them.