Google AI Search

You run an MCP server that exposes Google AI-powered search and documentation tools to help developers bridge knowledge gaps in real time. This server connects your client workflows to powerful models and web-grounded information, letting you query, analyze, and document programming topics with structured tooling.

How to use

You run the MCP server as a local or remote service and connect to it through an MCP client or integration layer. Once started, you can invoke tools such as web search, topic explanations with docs, and code/documentation retrieval to perform real-time information grounding, produce code snippets, and generate structured guidelines. Start the server with your preferred method (node script, or package-runner) and ensure your client is configured to target the active MCP endpoint. Use the available tools to ask questions, fetch documentation, and compare technologies with evidence-backed results.

How to install

Prerequisites you need before installing and running the MCP server: you must have Node.js v18 or newer and Bun installed. You also require credentials or API keys if you plan to use Vertex AI or Gemini integration.

# Install Bun (if not already installed)
# Bun is used for dependency management and running scripts
bun --version

# Install dependencies for the MCP server
bun install

# Build the server (compiles TypeScript to JavaScript)
bun run build

# Prepare environment (examples, you will tailor these in your .env)
# AI_PROVIDER can be "vertex" or "gemini"
# If using Vertex AI, provide GOOGLE_CLOUD_PROJECT and location as needed
# If using Gemini, provide GEMINI_API_KEY

# Start the server locally (examples shown in usage sections)

Configure environment variables in a file named ".env" at the project root, or supply them in your startup command. The required variables depend on your chosen provider.

Additional configuration notes

You can configure the MCP server by specifying environment variables to control model selection, streaming behavior, token limits, and retry logic. By default, streaming is enabled and minimal safety filtering is applied to reduce blocking while maintaining usable safeguards.

If you want to run the server using Node directly from a built path, you can reference the generated entry point after building.

Development

For active development, you can use watch and inspector tooling to iterate on the MCP server more quickly.

{
  "mcpServers": {
    "google-ai-search-mcp": {
      "command": "node",
      "args": ["/full/path/to/your/google-ai-search-mcp/build/index.js"],
      "env": {
        "AI_PROVIDER": "vertex",
        "GOOGLE_CLOUD_PROJECT": "YOUR_GCP_PROJECT_ID",
        "GOOGLE_CLOUD_LOCATION": "us-central1",
        "AI_TEMPERATURE": "0.0",
        "AI_USE_STREAMING": "true",
        "AI_MAX_OUTPUT_TOKENS": "65536",
        "AI_MAX_RETRIES": "3",
        "AI_RETRY_DELAY_MS": "1000",
        "GEMINI_API_KEY": "YOUR_GEMINI_API_KEY"
      },
      "disabled": false,
      "timeout": 3600
    }
  }
}

Available tools

answer_query_websearch

Developer-focused natural language queries with automatic technical detection, enhanced search methodology, and comprehensive code formatting using Google AI with real-time search results.

explain_topic_with_docs

Streamlined technical explanations that synthesize official documentation with enhanced debugging scenarios and troubleshooting guidance.

get_doc_snippets

Enhanced code snippet retrieval with progressive complexity, version targeting, and comprehensive context from official documentation.

generate_project_guidelines

Generates comprehensive structured project guidelines documents based on specified technologies using web search for current best practices.

code_analysis_with_docs

Evidence-based code analysis with citations, severity categorization, and actionable recommendations against official documentation practices.

technical_comparison

Technology comparison with quantitative benchmarks and evidence-based analysis across criteria.

architecture_pattern_recommendation

Architecture guidance with performance metrics, implementation roadmaps, and pattern recommendations.