Fantastic MCPs and where to find them
MCPs (Model Context Protocols) are creating new user behaviors, starting with developers. Let’s walk through where they’re working today and where they’ll expand next.
1. Giving coding agents up-to-date docs
MCPs first exploded in developer tools by solving a visceral pain point: AI agents hallucinating outdated API docs.
The Context7 MCP is one of the most popular ways of pulling docs for your agent. Add it to Cursor or VSCode, and your coding agent pulls real-time documentation for every library you use. No more manual checks or version mismatches. The agent now knows exactly how ai-sdk@2.3 works, not what it guessed from stale training data.
This pattern is now baked directly into many docs. You’ve likely spotted "Add to Cursor" or "Add to VSCode" buttons.
One-click installs a library’s docs MCP server directly into your IDE. No copy-pasting, no version drift. One fascinating side effect of this is that developers now trust agents to auto-add dependencies, knowing the agent always pulls the correct docs.
2. Giving human developers the right context
It’s intuitive to think of coding agents in the IDE as MCP clients. But the human developer who orchestrates and reviews generated code also needs equally good context.
Let’s say your agent pulls in a new framework you’ve never used. You don’t want to go read 10 blog posts to figure out what this framework does. You want a quick walkthrough of the framework right inside your codebase.
Mastra has a good solution to this that I'm starting to see spread. Mastra packages tutorials as MCPs you can trigger from your IDE with live code examples. No tab switching, no guesswork.
3. Taking actions on external resources
Agents are more helpful if they can touch the outside world. MCPs make that possible, starting with tools like GitHub.
The GitHub MCP lets agents (and you) create issues, review PRs, or triage bugs, all from the IDE. Same goes for project tools like Linear and Jira. Want to file a bug, assign a ticket, or draft a sprint plan? Just ask your agent.
The gap between planning and execution is now closed by MCPs.
Beyond coding: Where MCPs thrive
Developers, as usual, have been early adopters, but I expect MCPs to add value to any domain where two things are true:
- Workflows have high context gaps i.e. teams waste a lot of time switching tools or verifying stale data.
- Work happens in high-context-gravity clients like creation environments such as IDEs, Figma, or drafting tools.
We’re already seeing increased MCP adoption among GTM teams, where agentic CRMs use MCPs to pull live customer data from various sources, and marketers, where content creation tools use MCPs to wire up campaign metrics to specific assets.
I’m excited to see how this evolves.