Multi-Server MCP Config (All Five Presets Together)

Combining Multiple Servers

Every entry under mcpServers is a separate child process. Adding more servers does not change any other server's behavior — they're isolated by stdio.

Full config

{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-filesystem",
        "/Users/me/projects"
      ]
    },
    "github": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": { "GITHUB_PERSONAL_ACCESS_TOKEN": "ghp_xxx" }
    },
    "postgres": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-postgres",
        "postgresql://readonly:pw@localhost/dev"
      ]
    },
    "brave-search": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-brave-search"],
      "env": { "BRAVE_API_KEY": "BSA_xxx" }
    },
    "fetch": {
      "command": "uvx",
      "args": ["mcp-server-fetch"]
    }
  }
}

Startup cost

The client launches all five processes when it boots. With -y and cold npm caches that means a one-time 30-60 second wait on first run. Subsequent launches use the npm/uvx cache and start in 1-3 seconds per server.

Tool name conflicts

Each server exposes tools under its server name as a prefix (filesystem.read, github.search_issues, postgres.query), so two servers exposing similarly-named tools never collide. The model picks the right one from context.

Resource considerations

Each server is a Node.js or Python process holding ~50-150MB RAM. Five servers = roughly 500MB resident. Drop servers you're not actively using to keep the launch fast and your laptop battery happy.

Seeing what's loaded

In Claude Desktop, click the plug icon at the bottom-right of the input box to see which servers connected successfully and which failed. Failures usually mean missing env vars or wrong paths — see troubleshooting.