Claude Code MCP servers, explained

MCP (Model Context Protocol) is how Claude Code connects to external tools and data — databases, docs, APIs, design tools. Instead of pasting context by hand, an MCP server gives Claude a live, structured connection.

What people connect

Databases

Let Claude query and reason over your Postgres, MySQL or SQLite live.

Search & docs

Semantic code search or documentation retrieval to ground answers.

Issue trackers & repos

Read and act on GitHub, Jira or Linear from inside a session.

Design & browser

Pull from Figma or drive a browser for real UI context.

Internal APIs

Expose your own services so Claude can call them safely.

How to add one

MCP servers are configured per project or globally. Each server declares the tools and resources it exposes; Claude then calls them as needed. ConfigDeck lists ready-to-use MCP configs so you can add one without wiring it up from scratch.

Ready-made MCP configs

Browse Claude Code MCP configs and other setups on ConfigDeck.

Browse the marketplace →

FAQ

What command manages MCP connections?

`claude mcp` — with add, list, get, remove, login/logout and more.

How do I check if a server is connected?

Run `claude mcp list` — it health-checks each entry and shows Connected / Pending approval.

Where are connections stored?

In .mcp.json, project-level or user-level.

Should I connect every MCP server I find?

No — each connection widens the tool surface and permissions exposed; only connect what you actually use.