Press enter or click to view image in full size
AI coding agents can write code, fix bugs, create components, and even build entire applications.
But there’s something easy to overlook.
An AI agent can only work with the information and tools it has access to.
It might generate code using outdated documentation, struggle to understand your existing Figma design, or ask you to manually check whether a website actually works.
This is where MCP becomes interesting.
MCP (Model Context Protocol) lets AI applications connect to external tools and services. Instead of just generating suggestions, an agent can interact with things like your browser, GitHub repositories, databases, and design files.
Here are five integrations worth looking at.
Context7 — Give AI access to current documentation
One annoying problem with AI coding is outdated library knowledge.
Context7 helps by fetching documentation and relevant code examples for the libraries you’re working with.
Instead of relying entirely on what the model already knows, your agent can look up the documentation while working.
Playwright MCP — Let AI test your website
Generating a webpage is easy. Checking whether everything actually works is another matter.
Playwright MCP lets an AI agent interact with a real browser.
It can navigate pages, click buttons, fill forms, inspect content, and test workflows.
This is particularly useful when building a website and asking your agent to check its own work.
Figma MCP — Connect AI to your designs
This is an interesting one for frontend developers.
Get AJ’s stories in your inbox
Join Medium for free to get updates from this writer.
Figma’s MCP server lets coding agents access design context directly from Figma.
They can inspect layouts, understand components, retrieve design information, and, with supported workflows, write designs back to the canvas.
It makes the connection between design and code much more interesting.
GitHub MCP — Give AI access to your repositories
Your coding agent doesn’t have to work only with files on your computer.
GitHub’s official MCP server allows AI tools to interact with repositories, inspect files, search code, manage issues, and work with pull requests.
Useful when you want an agent to understand a project beyond the current file you’re editing.
Supabase MCP — Let AI work with your backend
If you build applications with Supabase, this one is worth exploring.
Its MCP server connects AI coding tools to your Supabase projects.
An agent can inspect database tables, write SQL queries, manage schema changes, work with migrations, and investigate project logs.
You can also configure read-only access and restrict the connection to a particular project.
That’s important because giving an AI agent access to your database also means thinking carefully about permissions.
The interesting thing about all this is that AI coding is slowly becoming less about writing better prompts.
It’s becoming more about giving agents the right context and tools to complete actual tasks.
You don’t need to connect everything at once. Start with the integration that solves a problem you already have.
I’d probably start with Context7 and Playwright for everyday frontend work.