Since January 1st 2026, the MCP connector has been available on the staging area of one of our important recent projects. This connector allows the system, which is a traditional management software product, to connect to an AI agent like ChatGPT or Claude, enabling users to operate by issuing requests in natural language instead of using the product’s conventional UI.
This represents an important milestone in the evolution of user interfaces for our systems, and the topic deserves further exploration.
The Evolution of User Interfaces
The advent of AI assistants like ChatGPT, Claude, and Gemini is inevitably changing how we interact with the web. The immediacy of responses from new deep agents, now a standard paradigm in all modern assistants, is becoming the preferred mode for our web searches.
Just as the spread of the web had a significant impact on user interface design for traditional applications in the past, it is easily foreseeable that the new conversational interfaces, which we are getting accustomed to in our daily interactions with AI assistants, will increasingly become the standard interface mode for any type of application.
From Chatbots to Agentic Systems
While the world of academic and industrial research ponders the future and evolution of LLM technologies, it is undeniable that companies engaged in the fierce competition unleashed in the sector continue to release products that have long surpassed the stage of wow-effect chatbot toys, beginning to be useful in everyday computer activities.
The turning point was the release of the first agentic applications, meaning software applications where the processing workflow is not predetermined by the programmer but is the result of an LLM’s output and “reasoning.”
New LLMs are trained to request the use of tools whenever the situation and assigned task requires additional information or the execution of particular actions.
A modern LLM can be invoked by passing in the input a description of a series of available software functions that can be called according to certain modalities. The description of these available functionalities becomes part of the prompt, and the LLM responds with a request to execute these functions if the assigned task requires it.
An agent is software that uses an LLM to analyze user input and understand their intent, to define an action plan, and to support every step of the plan until task completion.
Tools are the connecting element between the spoken reasoning of LLMs and the software applications we use daily.
An agent like ChatGPT is no longer just a content generator and an oracle to consult, but also a system capable of performing actions and activities by interfacing with other software products.
The MCP Protocol
In this regard, the rapid development of the MCP protocol is significant.
The Model Context Protocol (MCP) is an open standard that defines how AI models (especially LLMs) connect securely and uniformly to external tools, APIs, databases, and files, acting as a “translation layer” between assistant and real systems.
The official announcement of MCP as an open standard dates back to November 2024, with the release of SDK and open source reference servers.
The success was global and immediate. Since 2025, we have seen a growing ecosystem of MCP servers, registries, guides, and commercial/OSS tools.
AI providers and platforms are progressively adopting or supporting it, making it a de facto candidate as an industry standard for integration between agents and data/tools.
Fastal and MCP
At Fastal, we have been pioneers in its adoption, beginning experimentation since the first beta releases distributed by Anthropic, which is our main AI provider of reference.
In conjunction with the November 11, 2025 release, we decided to consolidate and release our framework for rapid development of MCP servers for management applications.
Starting from early 2026, new version releases of our legacy products and new custom systems will include MCP servers that will expose selected and controlled subsets of application functionality, enabling the creation of custom connectors on major commercial AI assistants.
Compatibility
Currently, Anthropic products Claude Web, Claude Code, and Claude Desktop fully support the MCP protocol and are completely compatible with our MCP servers.
OpenAI released in December 2025 in developers preview the custom MCP connector definition functionality and, with some limitations due to the still experimental nature of the feature on ChatGPT, the interface is compatible with our solutions.