3 MCP servers I actually use daily (and how to set them up)
Not a hype list. These are the three MCP servers I have running right now in Claude Desktop. 1. Filesystem mcp-hub install @modelcontextprotocol/server-filesystem I use this to let Claude read project files directly instead of me copy-pasting code. Configure it to point at your project root. Real usage : "Read the auth module and explain how the token validation works." 2. GitHub mcp-hub install…
Here are three MCP servers that I regularly use in Claude Desktop:
1. Filesystem mcp-hub install @modelcontextprotocol/server-filesystem
I configure this server to allow Claude to directly access project files, eliminating the need for me to copy-paste code. By pointing it to my project root, Claude can read the auth module and explain how token validation works.
Real usage: Check what changed in the last 3 PRs and summarize.
2. GitHub mcp-hub install @modelcontextprotocol/server-github
With this server, Claude can examine issues, pull requests, and browse repositories. I utilize it for code reviews by asking Claude to show me the changes made in recent PRs and provide a summary.
Real usage: Show me users who signed up this week but haven't logged in.
3. PostgreSQL mcp-hub install @modelcontextprotocol/server-postgres
This server enables Claude to directly interact with a PostgreSQL database, allowing me to run queries without leaving the conversation. I use it to retrieve insights, such as listing users who signed up in the past week but haven't logged in since.
Setup Instructions:
Each server takes approximately 2 minutes to set up:
Run the command: mcp-hub install server
Add the server to claude_desktop_config.json
Restart Claude Desktop
Important Note: The filesystem server can potentially read your entire disk if configured improperly. Exercise caution when specifying paths. All servers are verified using the npm package manager, and the mcp-hub CLI is available to discover additional options.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.