Think MCPs are just for code-whispering? Think again. Many developers are quick to dismiss MCPs (Model-Controlled Protocols) as just another IDE gimmick, especially after a less-than-stellar experience with coding agents. But here’s the part most people miss: MCPs are not just for developers.
This article, originally published on Block’s blog (https://block.github.io/goose/blog/2025/11/26/mcp-for-devs/), sheds light on the broader potential of MCPs. While Darren Shepherd’s tweet (https://x.com/ibuildthecloud/status/1990221860018204721) highlights a valid concern—many devs struggle to find value in MCPs for coding tasks—it’s crucial to understand that MCPs are a protocol, not just a tool. At Block, we’ve harnessed MCPs across finance, design, legal, and engineering, as demonstrated in my talk (https://youtu.be/IDWqWdLESgY?si=Mjoi-MGEPW9sxvmT). The key takeaway? MCPs are a foundation for building solutions that transcend traditional workflows.
But let’s focus on developers. If you’re skeptical, consider these dev-specific MCPs that might just change your mind:
GitHub MCP: Beyond the CLI
Yes, GitHub’s MCP (https://block.github.io/goose/docs/mcp/github-mcp) feels bloated right now, and the CLI is often faster for simple tasks. But here’s where it gets interesting: MCPs excel when your work spans multiple systems—GitHub, Slack, Jira—without you manually connecting the dots. Imagine resolving a bug, discussing fixes, and implementing changes all within a Slack thread, as our team did. No context switching, no browser tabs, just seamless collaboration. One team even had goose, our AI agent, complete 15 engineering days’ worth of work in a single sprint, leaving them scrambling for more tasks!
Context7: Documentation That Doesn’t Suck
Ever struggled with outdated docs or spent hours Googling API flows? Context7 MCP (https://block.github.io/goose/docs/mcp/context7-mcp) pulls real-time, up-to-date documentation, code examples, and guides directly into your AI agent. Ask it how to implement a Square SDK payment or Firebase auth flow, and get accurate, context-aware answers instantly. It’s like having a senior dev at your fingertips, without the interruptions.
Repomix: Map a Codebase in Minutes
Joining a new project? Repomix (https://block.github.io/goose/docs/mcp/repomix-mcp) compresses the entire codebase into an AI-optimized file, allowing you to query its structure, metrics, and patterns. Ask where the auth logic lives, how API calls work, or what’s still marked as a TODO. It’s like onboarding with a mentor who already knows the ins and outs, saving you hours of grepping and head-scratching. Even public GitHub repos are fair game—no cloning required.
Chrome DevTools MCP: Test While You Code
Frontend devs, rejoice! Chrome DevTools MCP (https://block.github.io/goose/docs/mcp/chrome-devtools-mcp) lets you test your web apps without leaving your IDE. Tell your agent to test a login form, and it’ll handle everything—opening Chrome, running tests, capturing screenshots, logging network traffic, and noting errors. It’s the ultimate productivity boost, saving you from writing custom scripts or manually clicking through tests.
The Bigger Picture
MCPs aren’t overhyped—they’re misunderstood. As Anthropic points out (https://www.anthropic.com/engineering/advanced-tool-use), the issue isn’t the protocol itself but how it’s used. Most setups overload the model with tools, leading to confusion. Goose, however, manages MCPs seamlessly (https://block.github.io/goose/docs/mcp/extension-manager-mcp), enabling and disabling them as needed. The real magic happens when you step outside the IDE and see how MCPs integrate AI into Slack, GitHub, Jira, Chrome, and more, creating workflows you never thought possible.
Controversial Question: Are developers underestimating MCPs by dismissing them as just another coding tool? Let us know in the comments—do you see MCPs as a game-changer, or are they still too niche for your workflow? And happy first birthday, MCP! 🎉
