• MATCH
  • RANK
  • CONTEXT
  • HUMAN

How to run vector similarity searches with your AI

Connect your AI to OpenSearch to find documents based on their actual meaning. Once connected, your agent can query your vector embeddings directly.

Ask AI about this page

Short answer

How do I run vector similarity searches with my AI?

Your AI uses the k-NN tool to compare the vector embeddings of your query against the vectors stored in your OpenSearch cluster. This allows your agent to pull up documents that share the same concept even if they don't use the same exact words. You'll see the most relevant results based on mathematical similarity.

Search outcomes

Where your results land

What happens when your agent queries the cluster.

MATCH

Semantic matches

The AI identifies documents that are conceptually close to your prompt. You get results that respect the intent behind your words.

RANK

Similarity scoring

The search tool ranks every document by its distance from your query. This ensures the most relevant data sits at the top of the list.

CONTEXT

Relevant data retrieval

Your agent pulls the specific text chunks needed to answer your question. This provides the raw material for your AI to process.

HUMAN

Review results

You look over the retrieved documents to ensure the similarity matches your specific needs. The AI presents the findings, but you decide if they are useful.

The workflow

What your AI does when a query arrives.

The AI handles the heavy lifting of vector math so you don't have to.

  1. Vectorize the query

    Your agent converts your plain text question into a high-dimensional vector that the cluster understands.

    search

  2. Execute k-NN search

    The AI sends a request to OpenSearch to find the nearest neighbors in the vector space.

    search

  3. Retrieve document hits

    The search returns the most similar document chunks along with their similarity scores.

    search

  4. Synthesize answers

    Your agent reads the retrieved text to formulate a coherent response to your original question.

    search

Try it

Copy these to start.

Use these prompts to trigger vector searches immediately.

Starting points

Replace the bracketed text with your specific index names or query topics.

OpenSearch Connector

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

Connector linkhttps://edge.vinkius.com/vk_preview_6R4dQ885yexHhsyH4M5e8fLiT1BjuMlzaUlJMumQ/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 — OpenSearch capabilities are ready to use.
Configuration · claude_desktop_config.jsonCopy
{
  "mcpServers": {
    "opensearch-vector-mcp": {
      "url": "https://edge.vinkius.com/vk_preview_6R4dQ885yexHhsyH4M5e8fLiT1BjuMlzaUlJMumQ/mcp"
    }
  }
}
Copy into chat04

  • Find documents in my 'customer_feedback' index that are conceptually similar to 'unhappy with shipping speeds'.

  • Search the 'product_manuals' cluster for information regarding battery maintenance and safety.

  • Run a k-NN search in my 'legal_docs' index to find clauses related to intellectual property rights.

  • Using the vector index 'tech_support_logs', find entries that match the concept of 'network connectivity issues'.

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

Start here

Connect OpenSearch Vector once, then ask.

Just provide your OpenSearch endpoint and credentials. We keep your login info encrypted, and you can start searching right away.

Connect OpenSearch Vector to your AI

FAQ

How the search behaves

  • 01

    Can the AI change my OpenSearch index settings?

    No. The AI only has permission to run searches. It cannot modify your cluster configuration or delete data.

  • 02

    Does the AI create new vector embeddings for me?

    No. This connector is for searching existing vectors. You must have already generated and stored your embeddings in OpenSearch.

  • 03

    Will the AI automatically update my index when I add new files?

    No. The AI only searches what is currently in the cluster. You handle the data ingestion process separately.

  • 04

    Can I use this to perform standard keyword searches?

    Yes, but the primary strength of this connector is performing k-NN vector similarity searches.

  • 05

    How does the AI know which index to search?

    You specify the index name within your prompt or conversation, and the AI uses that to target the correct part of your cluster.

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

Connect OpenSearch to Claude, Cursor, ChatGPT & more