You can now build and serve MCP Apps using Apollo Client and Apollo MCP Server without standing up or maintaining dedicated MCP infrastructure.
This release introduces a full framework and template for building conversational, interactive apps that run inside ChatGPT, Claude, Goose, Postman, and other MCP hosts. Instead of creating a new MCP Server per app or pushing that burden onto a platform team, frontend teams can now self-serve using familiar Apollo Client patterns.
Here’s what’s new:
- AI Apps Template (React + Vite + Apollo Client)
- Declarative
@toolsupport to turn GraphQL operations into MCP tools - Automatic MCP Tool + Resource generation via build artifacts
- One MCP Server can serve multiple MCP Apps
- Platform-specific component support using file-based conventions
- Protocol and metadata complexity abstracted away
With Apollo MCP Server 1.8.0, you deploy your app build artifacts next to a running MCP Server and it automatically serves your MCP Tools and Resources.
Release notes
https://github.com/apollographql/apollo-mcp-server/releases/tag/v1.8.0
Get started with the AI Apps Template
https://github.com/apollographql/ai-apps-template
Blog post
https://www.apollographql.com/blog/building-mcp-apps-with-apollo-client-and-apollo-mcp-server
See it in action
We’re hosting a walkthrough where we’ll demonstrate how to build and deploy an MCP App using Apollo Client and Apollo MCP Server, and explain how this fits into the evolving MCP ecosystem.
Register here: https://www.apollographql.com/events/building-mcp-apps-with-apollo-client-and-apollo-mcp-server
Heads up:
MCP Apps are part of a rapidly evolving ecosystem. While support across hosts is growing quickly, platform behavior and spec interpretation can still shift. You may encounter edge cases depending on the MCP host you’re targeting. We’re actively tracking feedback and continuing to improve compatibility.
We’d love to hear what you build and help with any issues you run into. Drop questions, feedback, or examples below 