andyshaman/gemini-webapi-mcp/skill

Gemini MCP Skill

Interact with Google Gemini via gemini-webapi-mcp MCP server.

Reference Router

How to Call

The tools below live on the gemini MCP server. Pick the path that matches how the server is registered — the tool names and arguments are identical in both:

• Registered in Claude Code (claude mcp add-json gemini …, see README) → the tools

are native: call mcp__gemini__gemini_generate_image etc. directly.

• Registered in [mcp-cli](https://github.com/philschmid/mcp-cli) (~/.config/mcp/mcp_servers.json)

→ run it from the shell; the JSON result carries images[].path and conversation_id: `` mcp-cli call gemini gemini_generate_image '{"prompt":"A red apple on a white table, 16:9"}' `` Image generation takes 20–60 s (up to several minutes for edits) — allow a long timeout.

Quick Start

Chat:

gemini_chat(prompt="Explain quantum computing")

Generate image:

gemini_generate_image(prompt="A cinematic wide shot of a futuristic city at sunset, volumetric fog, neon reflections on wet streets")

Edit image:

gemini_generate_image(prompt="Change the background to a sunset beach. Keep everything else exactly the same.", files=["/path/to/image.png"])

Analyze file:

gemini_upload_file(file_path="/path/to/doc.pdf", prompt="Summarize key points")

Analyze URL/YouTube:

gemini_analyze_url(url="https://youtube.com/watch?v=...", prompt="Summarize this video")

Key Facts

• Images saved to ~/Pictures/gemini/ as PNG, 2x upscaled resolution
• Watermark auto-removed (built-in, no extra dependencies)
• Include "16:9" or "9:16" in prompt for non-square aspect ratios
• Do NOT specify model for image generation — server picks the best one automatically
• Auth errors: call gemini_reset to refresh cookies
• Models and full parameter reference: see tools.md