# Vertex AI Search MCP > Vertex AI Search connects your agent directly to Google's semantic search engine, allowing you to ask complex questions about vast amounts of private company data. Instead of generic answers, it grounds responses in your own documents and knowledge bases. Manage structured datasets, find specific internal policies, or get personalized product recommendations—all through natural conversation. ## Overview - **Category:** ai-frontier - **Price:** Free - **Tags:** enterprise-search, grounding, semantic-search, generative-ai, natural-language-processing, data-retrieval ## Description This MCP lets your agent read and reason over your enterprise documentation like a human expert does. You connect it to any compatible AI client, and suddenly, your agent can stop hallucinating and start answering based on facts pulled from your own data stores. Need to know the current PTO policy? Or what the specs are for Product X? Instead of manually digging through shared drives or outdated wikis, you just ask your agent a question in plain language, and it pulls a direct, verifiable answer grounded in your internal documents. When you connect this MCP via Vinkius, you give your agent an entire knowledge layer built from scratch. You can even use the tool to list all available data sources so your agent knows exactly what information is accessible, making complex searches simple and repeatable. ## Tools ### get_grounded_answer Generates an answer in natural language using only information from your private documents. ### get_datastore_details Pulls the setup configuration and technical details for a specific data store. ### list_data_stores Lists all searchable document collections available in your Google Cloud project. ### list_datastore_documents Shows every indexed file or branch within a specified data store for review. ### list_search_engines Retrieves a list of all high-level search applications configured in the collection. ### get_recommendations Analyzes user behavior data to suggest relevant items or next steps for the user. ### search_documents Executes a general text search query across all documents in a specific repository. ## Prompt Examples **Prompt:** ``` List all my available data stores in Vertex AI Search. ``` **Response:** ``` I found 3 data stores: 'documentation-v1' (ID: doc-123), 'hr-policies' (ID: hr-456), and 'product-catalog' (ID: prod-789). Which one would you like to search through? ``` **Prompt:** ``` Based on our documentation, what is our remote work policy? ``` **Response:** ``` Grounded in the 'hr-policies' data store: Our remote work policy allows for up to 3 days of work from home per week, provided there is prior alignment with the team manager. Employees must ensure a stable internet connection and maintain core working hours. Would you like me to pull the full document? ``` **Prompt:** ``` Search the product catalog for 'blue wireless headphones'. ``` **Response:** ``` I found several matches in 'product-catalog': 1. 'CloudBass Pro Blue' (In Stock), 2. 'SkyBuds v2 Azure' (Limited Stock), and 3. 'Oceanic Beats Wireless' (Discontinued). I can provide more details on features or pricing for any of these. ``` ## Capabilities ### Ask questions using private documents It generates a natural language answer by retrieving and citing specific passages from your designated company documents. ### Identify available data sources You can list every searchable dataset or document collection you have configured within Google Cloud. ### Review data source configurations It retrieves specific metadata and setup details for any given data store, letting you check its status. ### Search across documents by query You perform a general search query against all indexed content within a specified document repository. ### Discover specific files and branches It lists every individual file or branch contained inside a target data store, helping you pinpoint sources of information. ### Get personalized product suggestions The agent retrieves recommendations by analyzing user interaction patterns against a specific dataset. ## Use Cases ### The HR team needs to update policy documentation. A manager asks their agent, 'What is the new parental leave policy?' The agent uses `get_grounded_answer` on the HR data store and replies with a direct quote and citation from the correct document version. ### Product teams are launching a new feature. An engineer asks, 'What features should we highlight for customers who bought Product A?' The agent uses `get_recommendations` on the product catalog data store and suggests related accessories or upgrades. ### A support rep needs to find a specific error code. The rep asks, 'Search for all mentions of error code 404b in our technical manual.' The agent uses `search_documents` and returns the exact document sections where that code is discussed. ### Data architects need to audit data sources. An architect asks, 'List all operational data stores.' The agent responds using `list_data_stores`, giving them a complete map of every available knowledge source for auditing purposes. ## Benefits - You eliminate guesswork. Instead of getting a generic answer from an LLM that might be wrong, you use the `get_grounded_answer` tool to ensure every piece of information comes directly from your verified internal documents. - Manage complex sources easily. Use `list_data_stores` and `get_datastore_details` to see exactly what datasets exist before you start querying them, saving time on failed searches. - Go deeper than keywords. The MCP performs semantic search across all content, meaning you don't have to know the exact terminology; just ask the question naturally. - Pinpoint sources of truth. If a document is misfiled or outdated, use `list_datastore_documents` to browse and see every indexed file inside a data store branch. - Understand user patterns. The `get_recommendations` tool lets your agent act like a personalized assistant by suggesting items based on past interactions. - View the full scope of search capabilities using `list_search_engines`, giving you an overview of all business-specific applications configured for searching. ## How It Works The bottom line is that you get reliable, fact-checked answers drawn directly from the knowledge base you already own. 1. Subscribe to this MCP and provide your Google Cloud Project ID, Location, and Access Token. 2. Your AI client authenticates the connection and establishes access to all your enterprise data stores. 3. You prompt your agent with a question or command (e.g., 'What is our remote work policy?'), and it uses its tools to retrieve and format an answer based solely on your documents. ## Frequently Asked Questions **How does Vertex AI Search MCP handle conflicting policies?** The agent is designed to prioritize grounded answers from your specific data stores. If conflicts exist across sources, it presents the findings and cites the source for you to resolve. **Do I need to pay extra for list_data_stores? ** No. Listing all available data stores is a foundational capability of this MCP and helps you map your entire knowledge footprint before you start querying. **Can Vertex AI Search MCP search live websites?** This MCP searches within the private documents and configured data stores you connect. It is not designed for general, real-time web crawling. **What if I want to find all mentions of a product ID?** You can use `search_documents` by providing the data store ID and the specific product ID query. This is far more effective than general searching. **How do I know what documents are available before connecting?** Use the `list_data_stores` tool first. It gives you a complete catalog of all searchable datasets, allowing you to understand your data scope.