MartinoPolo/mp-gemini-fetch

Gemini CLI Fetch

Fetches content from websites that Claude cannot directly access using Gemini CLI as a fallback.

Prerequisites

• Gemini CLI must be installed: npm install -g @anthropic/gemini-cli or equivalent
• Gemini must be authenticated (run gemini auth first)

Usage

Invoke with a URL:

/mp-gemini-fetch https://reddit.com/r/programming/top

Workflow

Step 1: Validate URL

Check if the URL is provided and valid.

Step 2: Execute Gemini CLI

IMPORTANT: Use -p flag for non-interactive mode. Without it, Gemini expects stdin and fails.

gemini -p "Fetch and summarize the main content from this URL: [URL]. Include key points, any code snippets, and relevant discussion highlights."

Use longer timeout for complex pages:

gemini -p "..." --timeout 120000

Step 3: Capture Output

Parse the Gemini response and format it for use.

Step 4: Return Results

Present the fetched content to the user:

"Fetched from: [URL]

Content: [Gemini's summary/content]

Note: This content was fetched via Gemini CLI as a fallback."

Common Use Cases

• Reddit threads (blocked by default)
• Sites with aggressive bot detection
• Content behind soft paywalls
• Forums and discussion boards

Error Handling

If Gemini CLI fails:

1. "No input provided via stdin" → You forgot the -p flag. Use gemini -p "prompt"
2. Check if Gemini is installed: gemini --version
3. Check authentication: gemini auth
4. Try with simpler prompt
5. Report error to user with troubleshooting steps

Notes

• This is a fallback for sites Claude cannot access directly
• Gemini CLI must be installed and authenticated separately
• Response quality depends on Gemini's access to the site
• Some sites may still be inaccessible