Skip to main content
Glama

gemini-understanding

A deliberately small MCP server exposing Gemini's web and multimodal understanding through five tools:

  • web

  • video

  • audio

  • document

  • image

The server uses the Gemini Interactions API. It does not inspect content to guess its modality. The calling agent chooses the tool, and Gemini returns the provider-side error when a modality, MIME type, URL, model, or thinking-level combination is invalid.

Tool behavior

web

Input:

{
  "prompt": "Research the latest development and also read https://example.com/page"
}

Both Gemini server tools are enabled on every call:

  • Google Search

  • URL Context

Use this for general web research and ordinary webpages. Use one of the typed tools for direct media/file URLs.

video, audio, document, and image

All four tools share this input:

{
  "source": "a local path or public HTTP(S) URL",
  "prompt": "What should Gemini do with it?",
  "mimeType": "optional MIME type hint"
}

Examples:

{
  "source": "https://www.youtube.com/watch?v=...",
  "prompt": "Summarize this video"
}
{
  "source": "https://example.com/video.mp4",
  "prompt": "Describe the video",
  "mimeType": "video/mp4"
}
{
  "source": "/workspace/report.pdf",
  "prompt": "Summarize the conclusions"
}

For remote sources, the URL is passed directly as a typed Gemini URI. YouTube is handled by calling video and normally omitting mimeType.

For local sources, the server:

  1. resolves the path relative to the MCP server's working directory;

  2. uploads it with the Gemini Files API;

  3. waits for processing;

  4. sends the uploaded URI as the caller-selected modality;

  5. deletes the temporary upload.

The SDK infers local MIME types from file extensions when possible. The server performs no MIME sniffing, URL header probing, extension-based modality routing, or automatic fallback to another tool.

Related MCP server: gemini-mcp

Configuration

GEMINI_API_KEY          required
GEMINI_MODEL            default: gemini-3.5-flash-lite
GEMINI_THINKING_LEVEL   default: high

Custom model and thinking-level strings are passed through without compatibility validation.

Install from GitHub

claude mcp add gemini-understanding -s user -- \
  env GEMINI_API_KEY=YOUR_KEY \
  npx -y github:sandlong/gemini-understanding

With explicit optional settings:

claude mcp add gemini-understanding -s user -- \
  env GEMINI_API_KEY=YOUR_KEY \
      GEMINI_MODEL=gemini-3.5-flash-lite \
      GEMINI_THINKING_LEVEL=high \
  npx -y github:sandlong/gemini-understanding

Local development

npm install
npm test
npm run build

Run the stdio server:

GEMINI_API_KEY=YOUR_KEY npm start

Error behavior

The server does not silently correct or reroute failed requests. Tool errors include the tool name, processing stage, configured model, thinking level, source where applicable, and the original Gemini status/code/message when available.

Local paths that do not exist fail at the filesystem stage because no API request can be constructed. Temporary-upload cleanup failures are non-fatal and are reported as warnings.

A
license - permissive license
-
quality - not tested
C
maintenance

Maintenance

Maintainers
Response time
Release cycle
Releases (12mo)
Commit activity

Resources

Unclaimed servers have limited discoverability.

Looking for Admin?

If you are the server author, to access and configure the admin panel.

Related MCP Servers

View all related MCP servers

Related MCP Connectors

  • MCP server exposing the Backtest360 engine API as tools for AI agents.

  • MCP server for Google Veo AI video generation

  • MCP server for AI dialogue using various LLM models via AceDataCloud

View all MCP Connectors

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/sandlong/gemini-understanding'

If you have feedback or need assistance with the MCP directory API, please join our Discord server