The Model Context Protocol's impact on 2025

Although the Model Context Protocol (MCP) was launched in November 2024, it would be hard to provide a convincing snapshot of technology in 2025 without discussing its incredible rise over the last 12 months. An open-source standard that helps developers connect AI systems with external applications and data sources, it’s played a significant role in driving AI adoption and experimentation forward. It’s arguably brought agentic AI into the mainstream much faster than the industry may have expected — by making it easier for developers to connect agents to many different sources of data, it's now possible to provide agentic systems with detailed and rich detailed context than would otherwise be possible without significant time and investment.

It shouldn’t be surprising, then, that MCP featured on the most recent volume of the Technology Radar (in Platforms/Trial). However, to really understand its impact it needs to be viewed as a force at the center of a growing ecosystem of tools and MCP servers. As we note in the publication, there’s a lot of innovation happening around MCP, and it ranges from major players getting involved, such as JetBrains, to tiny independent open source projects. 

In turn, this is shaping the way we build software too. It’s hard, for example, to separate the term context engineering from MCP. But this isn’t to say MCP is the silver bullet for AI adoption; there are security challenges and antipatterns emerging, as happens with any hyped technology. In this blog post we’ll take a look at MCP's impact on the year through the lens of Technology Radar Vol.33.

The proliferation of MCP servers

To get to grips with MCP’s influence, one the best places to begin is with MCP servers. MCP servers are specific programs that make it possible to connect AI models to other applications without requiring the usual — and often tricky — configuration and handling that developers would normally have to do. 

There are currently tens of thousands of MCP servers available, all for different tasks, challenges and tools. Many are curated and searchable on marketplace-style directories, such as MCP.so. And although some MCP servers are built by major players in the industry and tied to specific products or tools, the reason the space is so vibrant is that people are building their own MCP servers to tackle common challenges. This is one of the reasons we put FastMCP (Languages and Frameworks/Trial) on Technology Radar Vol.33 — it’s a Python framework that simplifies MCP server development.

One MCP server we did flag on the November 2025 Technology Radar is Context7 (Tools/Trial), which addresses inaccuracies in AI-generated code by providing LLMs with up-to-date, version-specific documentation and code examples. As we note in the publication, we’ve found Context7 to be very useful in AI-assisted software development. Helpfully, it can also be configured with a number of different code editors, including Claude Code and Cursor.

New techniques enabled by MCP

The adoption of MCP has gone hand-in-hand with a number of new techniques. Perhaps the most important to note is context engineering (Techniques/Assess), something we’ve been discussing a lot at Thoughtworks. Context engineering is, as we write in the Radar, about “the systematic design and optimization of the information provided to a large language model.” This involves grappling with a range of things, from prompts to memory to data retrieval. While MCP is only a part of this picture, it's undoubtedly an important one. The C in the middle reminded technologists of how critical context is when it comes to AI systems, opening up a space for context engineering to enter the conversation.

Elsewhere, we’ve also noticed the evolution of AI-powered UI testing (Techniques/Assess) this year; major UI testing tools such as Playwright and Selenium have introduced their own MCP servers (playwright-mcp and mcp-selenium), which bring more reliability to such tests.

Finally, it’s also worth calling out a technique we’ve been exploring for agentic coding: Anchoring coding agents to a reference application (Techniques/Assess). It addresses the age-old problem of code drift, where the live state of an application differs from how it's defined in code. It’s easy to see how such an issue could prove particularly troublesome for AI agents — by employing an MCP server to help anchor agents to template code and commit diffs, it becomes easier for those agents to detect and mitigate drift.

MCP risks and antipatterns

As with any rapidly adopted and much–hyped technology or trend, MCP isn't without risks. The most significant is security. As one widely shared article joked, the S in MCP stands for security. The piece, by researcher Elena Cross, outlines a number of common attack vectors opened up by MCP. This includes tool poisoning, where the MCP tool contains a malicious description, silent or mutated definitions and cross-server tool shadowing, where a malicious agent intercepts calls made to one that’s trusted. She makes the point that the protocol’s focus is on simplicity and ease, not authentication and encryption. 

On the recent Technology Radar we featured toxic flow analysis for AI (Techniques/Assess), a technique that seeks to map the various ways in which data flows through an agentic system, and identify vulnerabilities at different points of interaction. We think this is going to become a critical practice as the use of AI agents increases; as such, we're happy to see tools such as MCP-scan (Tools/Assess) emerge on the scene to facilitate these kinds of investigations.

While there are undoubtedly technical risks associated with MCP, some caution about when and where to use MCP could go a long way to mitigating many issues. For instance, we’ve noticed a rush to convert APIs to MCP servers. This is a trend that raises serious issues from both a security and efficiency perspective, which is why we’ve urged caution against what we describe as naive API-to-MCP conversion (Techniques/Hold) on Technology Radar Vol.33.

Looking ahead: MCP in 2026

Some of the most exciting developments in technology this year have been driven by MCP. As already mentioned, it’s difficult to imagine the industry will be in the position it is currently when it comes to agentic systems were it not for the emergence of the protocol. And while it was introduced by Anthropic,  it's the bottom-up innovation that's really giving the ecosystem energy.

If we can combine a thriving ecosystem and developer enthusiasm with maturing practices, we will doubtless see an interesting and innovative 2026.