Developers lose the focus 1200 times a day – how MCP could change that

Software developers spend most of their time not Write code; Recent industrial research has shown that actual coding explains as little as 16% of the working hours of developersWith the rest consumed by operational and supporting tasks. Because technical teams are under pressure to “do more with less” and CEOs brag about how much of their codebase is written by AI, a question remains: what was done to optimize the remaining 84% of the tasks that engineers work on?

Hold developers where they are most productive

An important culprit for the productivity of developers is contextom switching: the constant jumping between the ever -growing series of tools and platforms that are needed to build and send software. A Harvard Business Review research showed that the average digital employee browses almost between applications and websites 1200 times a day. And every interruption is important. The University of California discovered that it takes about 23 minutes To fully regain the focus after a single interruption, and sometimes worse, as almost 30% of Interrupted tasks are never resumed. Context exchange is actually central in Dora, one of the most popular frameworks for developing software for performance.

In an era in which AI-driven companies try to do their employees more to do more with less, further than “just” to give them access to large language models (LLMS), some trends are on the rise. For example, Jarrod Ruhland, chief engineer at Brex, assume That “developers deliver their highest value when they concentrate in their integrated development environment (IDE)”. With that in mind he decided to find new ways to make this happen, and the new protocol from Anthropic is perhaps one of the keys.

MCP: a protocol to bring context to ides

Coding assistants, such as LLM-driven Idos such as Cursor, Copilot and Windsurf, are in the middle of a Developer Renaissance. Their adoption speed is unseen. Cursor became the fastest growing Saas in history, with $ 100 million arrem within 12 months after launch, and 70% of Fortune 500 Companies use Microsoft Copilot.

But these coding assistants were only limited to codebase context, who could help developers write code faster, but could not help switching context. A new protocol deals with this problem: Model Context Protocol (MCP). Published by Anthropic in November 2024, it is an open standard that has been developed to facilitate integration between AI systems, in particular on LLM-based tools, and external tools and data sources. The protocol is so popular that there has been one 500% increase of new MCP servers in the past 6 months, with an estimated 7 million downloads in June,

One of the most impactful applications of MCP is the ability to connect AI coding assistants directly to the tools that developers depend on every day, workflows streamlined and the switching of context is drastically reduced.

Take function development as an example. Traditionally it is bouncing between different systems: reading the ticket in a project tracker, looking at a conversation with a teammate for clarification, the search for documentation for API details and finally, opening the IDE to start coding. Each step lives in a different tab, where mental shifts are needed that delay developers.

With MCP and modern AI assistants such as Claude van Anthropic, that entire process can take place in the editor.

For example, the implementation of a function of a function within a coding assistant becomes:

The same principle can apply to many other engineering workflow, for example an incident response for SREs can look like:

Nothing new under the sun

We have seen this pattern before. In the past decade, Slack has transformed the productivity of the workplace by becoming a hub for hundreds of apps, allowing employees to manage a wide range of tasks without leaving the chat window. Slack’s platform reduced context switching in daily workflows.

Riot games, for example, connected around 1,000 weak apps, and engineers saw one 27% Reduction In the time needed to test and repeat code, a 22% faster time to identify new bugs and an increase of 24% in the launch rate of the position; All were attributed to streamlining workflows and reducing the friction of tool switches.

Now a similar transformation takes place in software development, with AI assistants and their MCP integrations that serve as a bridge for all these external tools. In fact, the IDE could become the new all-in-one command center for engineers, just as Slack has been for general knowledge workers.

MCP may not be ready for Enterprise

MCP is a relatively emerging standard, for example, security wisem MCP has no built-in authentication or authorization model, which depends on external implementations that are still evolving, there is also ambiguity about identity and auditing-the protocol does not clearly distinguish whether an action is checked by a check and access and access and access and access and access and access and access to it and access and access to it, and access and access to it, and accessed and accessed and accessed with an additional one or the AI Custom solutions. Lori Macvittie, prominent engineer and main evangelist in the F5 Networks’ Office of the CTO, out That MCP “breaks the assumptions of the core protection that we have kept for a long time.”

Another practical limitation arises when too many MCP tools or servers are used at the same time, for example within a coding assistant. Every MCP server advertises a list of tools, with descriptions and parameters, which should consider the AI ​​model. The model flooding with dozens of available tools can overwhelm its context window. The performance degrades noticeable as the number of tools grows with some IDE integrations, have imposed hard limits (around 40 tools in cursor IDE, or ~ 20 tools for the OpenAI agent) to prevent

Finally, there is no refined way to be automatically discovered or presented contextually, except for mentioning them all, so developers often have to manually switch or put together which tools are active to make things work smoothly. Referring to that example of Riot games that install 1,000 weak apps, we can see how it can be unsuitable for the use of companies.

Less swivel chair, more software

Over the past decade, we have taught us the value to bring work to the employee, from slack channels that in updates to “Inbox Zero” e -mail methods and Unified Platform Engineering Dashboards. Now, with AI in our toolkit, we have the opportunity to enable developers to be more productive. Suppose Slack became the hub of business communication.

In that case, coding assistants are well positioned to become the hub of software recreation, not only where code is written, but where all context and employees come together. By keeping developers in their stream, we remove the constant mental equipment that has plagued the productivity of technical engineering.

Look for every organization that depends on software delivery, how your developers spend their day; You may be surprised by what you think.

Sylvain Kalache trains AI Labs Rooty.