• MATCHES
  • CONTEXT
  • DATA
  • HUMAN

How to perform semantic vector searches with your AI

Connect your agent to Google Cloud once. From then on, it can query live vector endpoints to find data based on meaning rather than just keywords.

Ask AI about this page

Short answer

How can I run semantic similarity searches with my agent?

Your agent uses specific embedding arrays to query live Google Cloud vector endpoints. It identifies the closest matches in your dataset through similarity lookups. You get precise, context-aware results.

Search outcomes

Where your results land.

What happens when your agent runs a lookup.

MATCHES

Nearest neighbor results

The agent identifies the most similar data points in your vector index. You receive the specific records that match your query's intent.

CONTEXT

Semantic relevance

The search ignores exact keyword matches in favor of conceptual similarity. This ensures your agent finds the right information even if the wording differs.

DATA

Live endpoint access

The agent pulls directly from your active Google Cloud endpoints. You work with the most current embeddings in your vector database.

HUMAN

Manual review needed

The agent presents the top matches for you to inspect. You decide if the similarity score meets your specific requirements.

The workflow

What your AI does when the query arrives.

The agent handles the heavy lifting of vector math and endpoint communication.

  1. Vectorize the query

    Your agent prepares the input to match the format required by your Google Cloud index.

    search_nearest_neighbors

  2. Execute lookup

    The agent sends the embedding array to the live endpoint to find matches.

    search_nearest_neighbors

  3. Retrieve neighbors

    The system pulls the closest data points based on the similarity score.

    search_nearest_neighbors

  4. Present results

    The agent summarizes the found neighbors so you can act on them.

    search_nearest_neighbors

Try it

Copy these to start.

Use these prompts to trigger a search.

Starting points

These are opening lines. Swap the bracketed text for your actual endpoint IDs and embedding arrays.

Vertex AI Vector Search Connector

You're all set. Choose your MCP client and follow the setup instructions.

Connector linkhttps://edge.vinkius.com/vk_preview_7qmxXVVjzgLzy5Qveg1Au3aDsjeacjAMcHp2XBIt/mcp

Claude Desktop

Follow the steps below to connect in seconds.

  1. 1In Claude Desktop, open Settings → Connectors.
  2. 2Click “Add custom connector” and paste the connector link above as the remote MCP server URL.
  3. 3Click Add and start a new chat — Vertex AI Vector Search capabilities are ready to use.
Configuration · claude_desktop_config.jsonCopy
{
  "mcpServers": {
    "vertex-ai-vector-search-mcp": {
      "url": "https://edge.vinkius.com/vk_preview_7qmxXVVjzgLzy5Qveg1Au3aDsjeacjAMcHp2XBIt/mcp"
    }
  }
}
Copy into chat04

  • Search for the nearest neighbors in my Google Cloud index using this embedding array: [0.1, 0.2, 0.3...]

  • Find similar items in my vector endpoint for the concept of 'customer churn' using this array: [0.5, -0.1, 0.8...]

  • Run a similarity lookup against my live endpoint using this specific embedding: [0.9, 0.4, -0.2...]

  • Identify the closest data points to this vector in my Vertex AI index: [0.0, 1.1, -0.5...]

  • Claude
  • ChatGPT
  • Cursor
  • VS Code
  • Windsurf
  • Claude Code
  • JetBrains
  • Cline

Start here

Connect Vertex AI Vector Search once, then ask.

Just link your Google Cloud credentials. Your login stays encrypted on our side, and you can start querying immediately.

Connect Vertex AI Vector Search to your AI

FAQ

How this task behaves

  • 01

    Can the AI update my vector index?

    No. This connector only performs similarity lookups; it cannot add, delete, or modify the data within your Google Cloud index.

  • 02

    Do I need to provide the embedding arrays?

    Yes. The agent requires specific embedding arrays to execute the search against your endpoints.

  • 03

    Can the AI change my Google Cloud permissions?

    No. The agent only uses the permissions you have already established for the endpoint.

  • 04

    Does the AI decide which matches are 'correct'?

    No. The agent identifies the mathematically closest neighbors, but you must review the results to determine if they are contextually useful.

  • 05

    Can I search by text instead of vectors?

    No. This specific task is designed for semantic vector lookups using embedding arrays.

  • More questions about Vertex AI Vector Search? The Connector page answers them. See everything the Vertex AI Vector Search Connector can do

Connect Vertex AI Vector Search to Claude, Cursor, ChatGPT & more