Software developers primarily spend their time on activities other than coding. Recent industry research indicates that coding comprises only 16% of their hours, with most time dedicated to operational and supportive tasks. As engineering teams face pressure to enhance productivity, CEOs now boast about AI contributions to their codebases, prompting questions about optimizing the remaining 84% of tasks engineers handle.
### Keep Developers Where They’re Most Productive
A significant factor in reducing developer productivity is context switching, where developers frequently shift between tools and platforms needed for software development. A Harvard Business Review study reveals that digital workers switch between applications nearly 1,200 times daily, with each interruption taking about 23 minutes to regain focus fully. This phenomenon impacts task resumption, and context switching is central to DORA, a key performance framework.
In this environment, businesses are harnessing AI to empower employees beyond just providing access to large language models (LLMs). Jarrod Ruhland, principal engineer at Brex, suggests that developers are most valuable when focused within their integrated development environment (IDE). Innovations like Anthropic’s new protocol aim to facilitate this focus.
### MCP: A Protocol to Bring Context to IDEs
The rise of coding assistants, such as LLM-powered IDEs like Cursor, Copilot, and Windsurf, marks a developer renaissance, quickly adopted by companies like Microsoft. However, these tools have been limited to codebase context. To address this, the Model Context Protocol (MCP), released in November 2024 by Anthropic, enables integration between AI systems and external tools, significantly reducing context switching.
MCP connects AI coding assistants directly to essential tools, streamlining workflows and minimizing the need for context switching. Traditional feature development requires navigating various systems for tasks, but with MCP, the entire process can occur within the IDE. Similarly, incident responses for SREs could also be integrated into this workflow.
### Nothing New Under the Sun
This transformation mirrors past advancements like Slack, which became a hub for workplace productivity, reducing the friction of tool-switching. For example, Riot Games’ integration of Slack apps led to significant productivity improvements. Similar changes now occur in software development, with AI assistants and MCP integrations becoming central to the workflow.
### MCP May Not Be Enterprise Ready
Despite its potential, MCP is still evolving. It lacks built-in authentication and permission models, posing security challenges. The protocol struggles with identity and auditing, complicating accountability and access control. The protocol’s performance also suffers when overloaded with too many tools, calling for hard limits in integration. Additionally, tool discovery remains manual, impairing seamless usage at scale.
### Less Swivel-Chair, More Software
The emphasis is on reducing context switching by keeping developers in their flow. Coding assistants could become the central hub for software creation, integrating context and collaboration within a cohesive platform. Organizations relying on software delivery should evaluate how their developers spend their time to identify productivity improvements.
Sylvain Kalache leads AI Labs at Rootly.