Gemma Ai Mcp Server

@rocnubie

13 days ago

Overview

Gemma AI MCP Server

Gemma AI - Free Google Gemma Chat with Advanced AI Models

A Model Context Protocol server that exposes the canonical Gemma AI knowledge surface — models, prompts, and chat workflows, pricing, FAQ, official links — to MCP-compatible AI clients such as Claude Desktop, Cursor, Windsurf, and Continue. Read-only, no API keys, no quota, ~50 ms cold start.

Official website: https://gemmaai.online

💬 About Gemma AI

Gemma AI (gemmaai.online) is a web-based platform that provides free access to Google DeepMind's Gemma 4 family of open-source AI models through a conversational chat interface. No credit card or subscription is required to get started. The platform covers four model variants spanning a wide capability range, from compact edge-optimized builds designed for mobile and browser deployment all the way to a flagship 31-billion-parameter dense model that ranks among the top three on the Arena AI public leaderboard. All models are released under the Apache 2.0 license, making them suitable for both personal experimentation and commercial use.

Key Features

Four-tier model family: Choose from the E2B/E4B ultra-compact edge models (2.3B and 4.5B effective parameters), a 26B Mixture-of-Experts variant that activates only 4B parameters per inference step, or the full 31B dense flagship — each tuned for different resource and performance trade-offs.
Native multimodal input: All models accept text, images at variable aspect ratios, video clips, audio, and documents (including OCR and diagram understanding) within a single conversation turn.
Extended context windows: Supports 128K to 256K token contexts with dual RoPE configurations, allowing long documents, codebases, and multi-turn sessions without truncation.
Strong benchmark performance: The 31B model scores 89.2% on AIME 2026 math reasoning, 80% on LiveCodeBench coding challenges, 85.2% on MMLU Pro, and a 2150 ELO rating on Codeforces — figures visible on the site's benchmark comparison table.
Flexible deployment paths: Beyond the hosted chat interface, models can run locally via Ollama, llama.cpp, or MLX; in browsers via transformers.js and WebGPU; and through ONNX checkpoints, Kaggle, or Hugging Face repositories.
Function calling without fine-tuning: Built-in support for autonomous agent workflows and tool integration, usable directly from the chat interface or via API.

Use Cases

Software development and competitive programming: Developers use the chat interface to generate, debug, and review code, with the 31B model achieving competitive-level performance on standard coding benchmarks.
Mathematical and logical reasoning: Students and researchers work through multi-step math problems, proofs, and quantitative analysis tasks that benefit from the model's high AIME and MMLU Pro scores.
Document and image analysis: Teams upload PDFs, screenshots, diagrams, or scanned pages to extract structured information, summarize content, or answer questions grounded in visual data.
Edge and privacy-sensitive applications: Developers building mobile or browser-based products use the compact E2B/E4B models, which run on-device without sending data to external servers.
Rapid prototyping with open-weight models: Researchers and startups evaluate Gemma 4 capabilities through the hosted interface before committing to local or cloud deployment under the permissive Apache 2.0 license.

Who Is It For

Gemma AI serves a broad technical audience. Developers and researchers who want access to frontier open-source models without a subscription barrier will find the free chat interface immediately useful. Teams evaluating models for commercial products benefit from the Apache 2.0 licensing and the range of deployment options. Students working on math, coding, or multimodal projects get access to high-performing models that would otherwise require cloud API budgets. Developers building privacy-first or offline-capable applications can use the edge-optimized variants as a starting point before moving to local deployment with Ollama, llama.cpp, or browser-native runtimes.

Tools

`list_models`

Return the canonical list of chat models exposed on the site, with capability notes. (Gemma AI)

Input: no parameters. Returns: text/markdown.

`get_pricing`

Return the canonical pricing entry point for Gemma AI.

Input: no parameters. Returns: text/markdown.

`get_official_links`

Return the canonical list of official links for Gemma AI (website, support, docs when available).

Input: no parameters. Returns: text/markdown.

Resources

site://gemmaai/models — Supported chat models and capability notes.
site://gemmaai/pricing — Canonical pricing entry point.
site://gemmaai/faq — Short FAQ generated from public site metadata.
site://gemmaai/links — Canonical URLs to share with users.

Prompts

`tell_me_about_gemmaai`

Summarize what the site is, who it's for, and how it works. — Gemma AI

`start_chat_session_gemmaai`

Open a chat-evaluation session against the site's models, with sensible defaults. — Gemma AI

Installation

Install via Smithery

npx -y @smithery/cli install gemmaai-mcp --client claude

(Replace claude with cursor, windsurf, or continue for those clients.)

Install from source

git clone https://github.com/rocnubie/gemmaai-mcp.git
cd gemmaai-mcp
pnpm install

Then add to your MCP client config (claude_desktop_config.json for Claude Desktop, mcp.json for Cursor / Windsurf / Continue):

{
  "mcpServers": {
    "gemmaai-mcp": {
      "command": "node",
      "args": [
        "/absolute/path/to/gemmaai-mcp/src/index.mjs"
      ]
    }
  }
}

Debug with MCP Inspector

npx @modelcontextprotocol/inspector node src/index.mjs

Official Links

Development

pnpm install
pnpm start                 # run the server over stdio

License

MIT

Try in Playground

Server Config

{
  "mcpServers": {
    "gemmaai-mcp": {
      "command": "node",
      "args": [
        "/absolute/path/to/gemmaai-mcp/src/index.mjs"
      ]
    }
  }
}

Build with ShipAny.