MCP Server for Vertex AI Search

What is MCP Server for Vertex AI Search?

MCP Server for Vertex AI Search is a Model Context Protocol server that searches documents using Vertex AI grounding with Gemini. It integrates one or multiple Vertex AI data stores to ground responses in private data.

How to use MCP Server for Vertex AI Search?

Clone the repository, create a virtual environment with uv, and install dependencies. Then run the server with uv run mcp-vertexai-search serve --config config.yml --transport <stdio|sse>. Alternatively, install the Python package from GitHub and use the same command. A YAML config file (template provided) is required.

Key features of MCP Server for Vertex AI Search

• Uses Gemini with Vertex AI grounding for accurate search
• Supports one or multiple Vertex AI data stores
• Offers both SSE and stdio transport modes
• Provides a dedicated search test command
• Can run locally or via Docker
• Configurable YAML configuration file

Use cases of MCP Server for Vertex AI Search

• Perform AI‑powered search over enterprise documents stored in Vertex AI
• Ground Gemini responses in private, organization‑specific data
• Build MCP clients that query multiple data stores simultaneously
• Test search quality directly without starting the full MCP server

FAQ from MCP Server for Vertex AI Search

What data sources does MCP Server for Vertex AI Search support?

It supports one or multiple Vertex AI data stores. The data stores are configured via a YAML config file.

What are the runtime dependencies?

Requires Python with uv (or pip) and access to Vertex AI (Gemini model and data store). A Google Cloud project with Vertex AI enabled is necessary.

What transports does the server support?

It supports SSE (Server‑Sent Events) and stdio (Standard Input Output), controlled by the --transport flag.

Is the package published to PyPI?

No, the package is not yet published to PyPI. It can be installed directly from the GitHub repository.

How to test the search without the MCP server?

Use the command: uv run mcp-vertexai-search search --config config.yml --query <your-query>.