Google Drive HPC Log Analyzer

You can run a local MCP server that connects Claude Desktop to your Google Drive and provides automated HPC log analysis. It lets you search, read, and analyze HPC logs directly from Google Drive without leaving Claude, helping you quickly identify errors, timeouts, and resource issues.

How to use

You will interact with the server through Claude Desktop after it is loaded as an MCP server. Start Claude Desktop, ensure the server is loaded (look for the 🔌 icon), then ask Claude to search for log files, read specific files, or run analysis on HPC logs. Typical tasks include finding log files across Drive, reading a selected log, and having Claude return actionable insights such as out-of-memory errors, time limit issues, failed jobs, and resource allocation problems.

How to install

Prerequisites: Python 3.7 or newer, Claude Desktop installed, a Google Cloud account, and pip for package management.

# Step 1: Clone the MCP server repository
git clone https://github.com/minitim222/gdrive_hpc_mcp.git
cd gdrive_hpc_mcp

# Step 2: Install dependencies
pip install -r requirements.txt

# Step 3: Set up Google Cloud credentials (credentials.json provided by Google Cloud Console)
# Place credentials.json in the project root or a secure location referenced by your setup

# Step 4: Configure Claude Desktop to load the MCP server
# You will generate an absolute path to the server script in the Claude config below

# Step 5: Start the MCP server (the server opens a browser for Google OAuth during first run)
python server.py

Configuration and operation notes

The server is designed to run locally and use read-only access to Google Drive. It relies on a one-time OAuth setup with token refresh handled automatically. Never commit credentials.json or token.pickle to public repositories.

To enable Claude Desktop to launch the server, provide the absolute path to the server script in Claude’s MCP configuration. Here is an example configuration you can adapt.

{
  "mcpServers": {
    "gdrive_hpc": {
      "command": "python",
      "args": ["/absolute/path/to/gdrive_hpc_mcp/server.py"]
    }
  }
}

Security and data access

Read-only access means the server requests permission to read files from Google Drive and does not modify or delete data. All authentication tokens are stored locally on your machine, and credentials.json and token.pickle should be excluded from version control. If you encounter authentication issues, you can re-authenticate by restarting the server and following the OAuth prompt.

Troubleshooting

If you see a credential error, ensure credentials.json is present and correctly referenced. For module or dependency issues, install dependencies again with pip install -r requirements.txt. If the MCP server does not appear in Claude Desktop, verify that the JSON configuration is valid and that the server path is absolute, then restart Claude Desktop.

Examples of use within Claude

• Search for log files in Google Drive
• Analyze the latest job log to determine why it failed
• Find all HPC logs from the past week
• Read a specific error log from your simulations folder

Available tools

list_drive_files

Search and list files in Google Drive with optional queries and folder filtering.

read_drive_file

Read the contents of a file from Google Drive.

analyze_hpc_log

Analyze an HPC log file and provide insights including errors, warnings, and actionable suggestions.

search_hpc_logs

Search specifically for HPC log files with extensions like .log, .out, or .err.