Diario del capitán, fecha estelar d627.y41/AB

The Model Context Protocol (MCP) community has introduced a new proposal called “MCP Apps” that aims to standardize how interactive user interfaces are built and shown inside AI assistants.
This matters because interactive components already existed in practice, but there wasn’t a common MCP-level standard for them.
Up to now, you could see interactive UI inside assistants, but each platform did it its own way:
The problem was simple: these approaches weren’t officially part of MCP itself, so developers had to adapt to different systems depending on the assistant or host.
The new MCP Apps proposal brings a shared direction: it defines an official way for MCP servers to send interactive UI blocks that any compatible assistant can render safely.
In other words:
Before: interactive components existed, but every host had its own “dialect”.
Now: MCP is moving toward one common standard so components can be built once and work across MCP clients.
For users, this is a step toward assistants that feel more like real apps, with buttons, forms, and visual mini-screens instead of text-only replies.
For developers, it removes fragmentation: no more rebuilding the same UI differently for each assistant, since MCP Apps aims to make interactive UI a first-class feature of the protocol.
The proposal is still in draft form, but it clearly signals where MCP is headed next: interactive AI experiences, standardized and portable.
At GPTApps, we’ll be adopting this new MCP Apps approach in our MCP server projects, so any client who wants to build an application for AI assistants can rely on a single, standardized way to deliver interactive components across hosts. If you’re exploring an MCP-based app or want help bringing your content and services to AI assistants like Gemini or ChatGPT, you can reach out via our contact page.

If your development provider holds the keys, switching can turn into a hostage situation. Here's how to make sure you truly own your product.
Leer el artículo: What we’ve learned rescuing projects from bad provider handovers
We’ve spent a few weeks with personal AI assistants that work in the background, connected to the tools we use to run MarsBased. The most useful thing so far: they flag what we didn’t ask them to look for.
Leer el artículo: What we’ve learned using always-on AI assistants
Decision models could make AI practical in applications where cost and complexity have kept it out. Here’s what they are and why we’re keeping an eye on them.
Leer el artículo: Decision models: practical AI where it used to be too costly