Popular MCP Servers

The MCP ecosystem includes dozens of ready-to-use servers. The ones below cover the most common integration needs — local file access, code repositories, databases, web search, and browser automation. All are open-source and installable via npm.

Filesystem Server

Package: @modelcontextprotocol/server-filesystem

Gives Claude read and write access to a directory on your local machine. You specify the allowed directory when starting the server — Claude cannot access files outside it.

{
  "mcpServers": {
    "filesystem": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/path/to/directory"]
    }
  }
}

Tools exposed: read_file, write_file, list_directory, create_directory, move_file, search_files. Use this when you want Claude to read documentation, edit config files, or work with data files in a specific project folder.

GitHub Server

Package: @modelcontextprotocol/server-github

Connects Claude to the GitHub API. Requires a personal access token with appropriate permissions.

{
  "mcpServers": {
    "github": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": { "GITHUB_PERSONAL_ACCESS_TOKEN": "ghp_your_token" }
    }
  }
}

Tools exposed: search repositories, read file contents, list and create issues, create and review pull requests, manage branches, get commit history. Useful for code review workflows, issue triage, and repository navigation.

Postgres Server

Package: @modelcontextprotocol/server-postgres

Connects Claude to a PostgreSQL database. Pass a connection string as the argument.

{
  "mcpServers": {
    "postgres": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-postgres", "postgresql://user:pass@localhost/dbname"]
    }
  }
}

Tools exposed: query (execute SQL and return results), list_tables, describe_table. Use for data analysis, schema exploration, and SQL generation with execution against real data. Point it at a read-only replica or dev database — not production with write access.

Brave Search Server

Package: @modelcontextprotocol/server-brave-search

Gives Claude web search capability via the Brave Search API. Requires a Brave Search API key (free tier available).

{
  "mcpServers": {
    "brave-search": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-brave-search"],
      "env": { "BRAVE_API_KEY": "your_api_key" }
    }
  }
}

Tools exposed: brave_web_search, brave_local_search. Use when you need Claude to research current events, look up documentation, or find information beyond its training data.

Puppeteer Server

Package: @modelcontextprotocol/server-puppeteer

Headless browser automation via Puppeteer. Lets Claude navigate web pages, take screenshots, click elements, and extract content from JavaScript-rendered pages.

{
  "mcpServers": {
    "puppeteer": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-puppeteer"]
    }
  }
}

Tools exposed: navigate, screenshot, click, fill, evaluate (run JavaScript), get_content. Requires Chrome or Chromium installed. Use for web scraping, UI testing workflows, and form automation.

Finding More Servers

The MCP ecosystem is growing rapidly. Places to find servers:

• Official reference servers: The modelcontextprotocol/servers GitHub repository contains Anthropic-maintained reference implementations including SQLite, Slack, Google Drive, Sentry, and more
• Community registry: mcp.so and glama.ai/mcp/servers aggregate community-built servers
• npm search: Search mcp-server on npmjs.com for JavaScript/TypeScript servers
• PyPI search: Search mcp on PyPI for Python servers

When evaluating a third-party MCP server: review the source code before running it, check what permissions it requests, and run it with minimal credentials until you trust it. MCP servers run as processes with your user permissions.

Checklist: Do You Understand This?

• Filesystem: local file read/write scoped to a directory — use for documents and project files
• GitHub: repository navigation, issues, and PRs — requires a personal access token
• Postgres: SQL execution and schema inspection — use a read-only replica for safety
• Brave Search: web search via Brave Search API — free tier available, needs API key
• Puppeteer: headless browser automation — navigate, screenshot, click, extract content
• Find more servers at the official modelcontextprotocol/servers repo and community registries