Need smarter insights in your inbox? Sign up for our weekly newsletter to get only the things that matter to enterprise AI, data and security leaders. Subscribe now
Software developers spend most of their time do not have Writing Code; Recent industry research shows that actual coding accounts for only 16% of developers’ working hours, with the rest being consumed by operational and supportive tasks. As engineering teams are under pressure to “do more,” CEOs boast about how much of the codebase is written by AI, what is being done to optimize the remaining 84% of the tasks engineers are working on?
Keep your developers in the most productive place
The main perpetrator to developer productivity is switching contexts. It’s a constant hopping between ever-growing tools and platforms needed to build and ship software. A Harvard Business Review survey found that the average digital worker flips between an application and a website nearly 1,200 times a day. And it’s important to interrupt everything. The University of California has discovered that it takes about 23 minutes to regain focus after a complete interruption, as almost 30% of suspended tasks are not resumed. Context switching is actually at the heart of Dora, one of the most popular performance software development frameworks.
Several trends are emerging at an age where AI-driven companies are trying to enable them to do more than have fewer employees and have access to “fair” and large-scale language models (LLMS). For example, Jarrod Ruhland, a leading engineer at Brex, assumes that “developers provide the best value when focusing on integrated development environments (IDEs).” With that in mind, he decided to find a new way to make this happen, and the new protocols of humanity may be one of the keys.
MCP: A protocol that brings context to IDES
Coding assistants such as LLM-driven IDEs such as cursors, copilots, and windsurfs are at the heart of the developer renaissance. Their adoption speed is invisible. Cursors have become the fastest growing SaaS in history, reaching a $100 million ARR within 12 months of launch, with 70% of Fortune 500 companies using Microsoft Copilot.
AI scaling reaches its limit
Power caps, rising token costs, and inference delays are rebuilding Enterprise AI. Join exclusive salons and discover what your top team looks like.
- Turning energy into a strategic advantage
- Architects efficient inference for real throughput gain
- Unlock competitive ROI with a sustainable AI system
Make sure you have your place to stay first: https://bit.ly/4mwgngo
However, these coding assistants were limited to codebase contexts only, which helps developers write code faster, but they didn’t help switching contexts. A new protocol addresses this issue: Model Context Protocol (MCP). Released by Anthropic in November 2024, it is an open standard developed to promote integration between AI systems, particularly LLM-based tools and external tools and data sources. The protocol is so popular that it has grown by 500% over the past six months, with an estimated 7 million downloads in June.
One of MCP’s most impactful applications is the ability to connect AI coding assistants directly to the tools they rely on every day. Streamline your workflow and dramatically reduce context switching.
Let’s consider functional development as an example. Traditionally, it involves bounce between several systems: reading tickets in the Project Tracker, looking at the conversation to clarify the conversation with your teammates, searching for documentation of API details, and finally opening the IDE to start coding. Each step is in a separate tab, which requires mental changes that slow the developer down.
With modern AI assistants like MCP and Anthropic’s Claude, the entire process can occur within the editor.
For example, to implement all the features within the coding assistant:
The same principles can be applied to many other engineer workflows. For example, an incident response for SRES would look like this:
There’s nothing new in the sun
I’ve seen this pattern before. Over the past decade, Slack has transformed workplace productivity 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 reduces context switching in everyday workflows.
For example, Riot Games has connected around 1,000 Slack apps, and engineers saw a 27% reduction in the time required to test and iterate the code, a 22% faster time to identify new bugs, and a 24% increase in feature firing rates. All was due to streamlining workflows and reducing tool switching friction.
Currently, the AI Assistant and MCP integration act as bridges to all these external tools, so similar conversions are occurring in software development. In fact, the IDE could become the new all-in-one command center for engineers, just as Slack is for the general knowledge worker.
MCP may not be ready for enterprise
MCP is a relatively early standard. For example, Wisem MCP for security does not have an authentication or authorization model built into it, so it relies on an external implementation that is still evolving. There is also ambiguity about identity and auditing. The protocol does not clearly distinguish whether the action was triggered or not by the user or the AI itself. Lori MacVittie, a well-known engineer and chief evangelist in F5 Networks’ CTO office, says MCP “breaks the core security assumptions we’ve held for a long time.”
For example, too many MCP tools or servers are being used simultaneously within the coding assistant, creating another practical limitation. Each MCP server advertises a list of tools with descriptions and parameters that the AI model should consider. With dozens of available tools, you can overwhelm the context window. Some IDE integrations result in significant degradation in performance as tool counts grow, imposing a stiff limit (approximately 40 tools in the Cursor IDE, or about 20 tools in the Openai agent) to prevent prompts from expanding beyond what the model can handle.
Finally, there is no sophisticated way to autodiscover or suggest contextually, as well as listing everything. Therefore, developers often switch manually or are active to which tools work actively. If you look at the example where Riot Games installs 1,000 Slack apps, you will see that it may be suitable for enterprise use.
Fewer swivel chairs and more software
The past decade has taught us the Inbox Zero email methodology and unified platform engineering dashboard the value of bringing work to workers through the updated slack channel. Now, when AI is featured in toolkits, there is an opportunity to empower developers to be more productive. Suppose Slack has become a hub for business communications.
In that case, coding assistants are well positioned not only to be where the code is written, but also to be a hub for software writing, where all the context and collaborators come together. By keeping developers in the flow, we remove certain mental gear shifts that are plaguing engineering productivity.
For organizations that rely on software delivery, take a closer look at how developers spend their day. You may be surprised at what you find.
Sylvain Kalache leads the AI lab at Rootly.
Source link
