Six templates, one minute to set up

1. 1.Scope filesystem access to a single project root.
2. 2.Generate a GitHub fine-grained token with repo + issues read.
3. 3.Optional: connect Sentry read token to triage production errors in-context.

{
  "mcpServers": {
    "github": {
      "env": {
        "GITHUB_TOKEN": "…"
      },
      "args": [
        "-y",
        "@modelcontextprotocol/server-github"
      ],
      "command": "npx"
    },
    "sentry": {
      "env": {
        "SENTRY_TOKEN": "…"
      },
      "args": [
        "-y",
        "@modelcontextprotocol/server-sentry"
      ],
      "command": "npx"
    },
    "filesystem": {
      "args": [
        "-y",
        "@modelcontextprotocol/server-filesystem",
        "/Users/you/code/project"
      ],
      "command": "npx"
    }
  }
}

1. 1.Install Claude Desktop and open MCP settings.
2. 2.Add the three server entries below to claude_desktop_config.json.
3. 3.Restart Claude. Ask: 'Summarize the AI model leaderboard changes this week.'

{
  "mcpServers": {
    "brave": {
      "env": {
        "BRAVE_API_KEY": "…"
      },
      "args": [
        "-y",
        "@modelcontextprotocol/server-brave-search"
      ],
      "command": "npx"
    },
    "fetch": {
      "args": [
        "-y",
        "@modelcontextprotocol/server-fetch"
      ],
      "command": "npx"
    },
    "memory": {
      "args": [
        "-y",
        "@modelcontextprotocol/server-memory"
      ],
      "command": "npx"
    }
  }
}

1. 1.Connect a read-only Postgres user to the database.
2. 2.Scope filesystem to a /data dir for CSV outputs.
3. 3.Ask Claude to 'Pull weekly active users by cohort and export to /data/wau.csv'.

{
  "mcpServers": {
    "postgres": {
      "args": [
        "-y",
        "@modelcontextprotocol/server-postgres",
        "postgresql://readonly@localhost/prod"
      ],
      "command": "npx"
    },
    "filesystem": {
      "args": [
        "-y",
        "@modelcontextprotocol/server-filesystem",
        "/Users/you/data"
      ],
      "command": "npx"
    }
  }
}

1. 1.Keep send-mode in 'confirm' until you trust the configuration.
2. 2.Add a whitelist of allowed recipient domains.
3. 3.Use memory to persist signatures and common templates.

{
  "mcpServers": {
    "email": {
      "env": {
        "SEND_MODE": "confirm",
        "SMTP_HOST": "…",
        "SMTP_USER": "…"
      },
      "args": [
        "-y",
        "mcp-server-email"
      ],
      "command": "npx"
    },
    "memory": {
      "args": [
        "-y",
        "@modelcontextprotocol/server-memory"
      ],
      "command": "npx"
    }
  }
}

1. 1.Generate a Linear API key with write scope for your team.
2. 2.Install the Slack server with a workspace OAuth token.
3. 3.Use memory to persist your team's canonical labels and cycle cadence.

{
  "mcpServers": {
    "slack": {
      "env": {
        "SLACK_TOKEN": "…"
      },
      "args": [
        "-y",
        "mcp-server-slack"
      ],
      "command": "npx"
    },
    "linear": {
      "env": {
        "LINEAR_API_KEY": "…"
      },
      "args": [
        "-y",
        "mcp-server-linear"
      ],
      "command": "npx"
    },
    "memory": {
      "args": [
        "-y",
        "@modelcontextprotocol/server-memory"
      ],
      "command": "npx"
    }
  }
}

1. 1.Share design Notion workspaces with the integration.
2. 2.Authorize Drive read scope for reference assets.
3. 3.Ask Claude to compile moodboards citing sources.

{
  "mcpServers": {
    "fetch": {
      "args": [
        "-y",
        "@modelcontextprotocol/server-fetch"
      ],
      "command": "npx"
    },
    "gdrive": {
      "env": {
        "GDRIVE_OAUTH": "…"
      },
      "args": [
        "-y",
        "mcp-server-gdrive"
      ],
      "command": "npx"
    },
    "notion": {
      "env": {
        "NOTION_TOKEN": "…"
      },
      "args": [
        "-y",
        "mcp-server-notion"
      ],
      "command": "npx"
    }
  }
}