# OpenSearch Vector MCP for AI Agents AI Agent Connect > OpenSearch Vector MCP lets you manage your vector database directly through your AI agent. You can create k-NN indexes, upsert embeddings with metadata, and run similarity searches without jumping between your terminal and the dashboard. It turns your OpenSearch cluster into a queryable knowledge base for RAG pipelines and ML projects. ## Overview - **Category:** loved-by-devs - **Price:** Free - **Endpoint:** https://edge.vinkius.com/vk_preview_6R4dQ885yexHhsyH4M5e8fLiT1BjuMlzaUlJMumQ/ai-agent-connect - **Tags:** vector-database, k-nn, search-engine, embeddings, indexing ## Description Most developers spend way too much time switching between their terminal, the OpenSearch dashboard, and their IDE just to check if an index is healthy or to see what a specific mapping looks like. This Connector changes that by letting your AI agent handle the heavy lifting. Instead of writing out long curl commands or manually constructing JSON payloads to upsert embeddings, you can just tell your agent what to do in plain English. It handles the connection to your cluster and performs the actions for you. Whether you are testing a new similarity query or need to quickly provision a new index for a production RAG pipeline, this tool puts the power of your vector store right into your chat interface. It is a huge time saver for anyone who needs to iterate on vector search logic quickly. You can find this and hundreds of other tools in the Vinkius catalog to keep your stack connected. ## Tools ### search Run a k-NN vector search against a specific index using a dense float vector array. ### list_indexes Get a full list of all explicit indexes currently residing on your OpenSearch cluster. ### get_index Retrieve the exact mapping and settings for a specific OpenSearch index. ### index_document Perform a fast, atomic insertion of a single vector document into your embedding space. ### delete_document Remove a specific vector document from your OpenSearch cluster by its ID. ### create_index Provision a new native OpenSearch KNN index optimized for cosine similarity. ## Prompt Examples **Prompt:** ``` List all my vector indexes and tell me which ones have a yellow health status. ``` **Response:** ``` Checking your OpenSearch cluster now... Found the following indexes: | Index Name | Status | Doc Count | Dimensions | | :--- | :--- | :--- | :--- | | **product-embeddings** | Green | 45,230 | 768 | | **support-tickets** | Green | 12,800 | 1536 | | **knowledge-base** | Green | 3,450 | 384 | | **test-index** | **Yellow** | 50 | 768 | The `test-index` currently has a yellow health status. Would you like me to inspect its settings? ``` **Prompt:** ``` Create a new index called 'user-embeddings' with 768 dimensions and cosine similarity. ``` **Response:** ``` I have successfully created the new index. **Index Details:** * **Name:** user-embeddings * **Dimensions:** 768 * **Similarity Metric:** Cosine * **Engine:** NMSLIB The index is now ready for you to start upserting documents. ``` **Prompt:** ``` Find the 3 most similar documents to a request about 'how to reset my password' in the support index. ``` **Response:** ``` Searching the `support-tickets` index for 'how to reset my password'... Top 3 results: 1. **Password Reset Guide** (Score: 0.98) 2. **Account Recovery Steps** (Score: 0.94) 3. **Security FAQ** (Score: 0.89) Would you like me to pull the full content of the top result? ``` ## Capabilities ### Run similarity searches Execute k-NN queries against your indexes to find the most relevant documents. ### Provision vector indexes Create new k-NN indexes with specific dimensions and similarity metrics. ### Upsert vector documents Add new embeddings and metadata to your existing vector space. ### Inspect index configurations View mappings, settings, and health status for any index in your cluster. ### Remove specific documents Delete individual vector documents from your OpenSearch cluster by their ID. ## Use Cases ### Testing RAG context A developer asks the agent to find the 5 most similar documents to a new query to see if the retrieval is accurate. ### Quick index provisioning An ML engineer needs a new 1536-dimension index for a new project and has the agent create it in one go. ### Health checks A data team member asks the agent to list all indexes to check for yellow health statuses during a migration. ### Metadata auditing A developer wants to see the specific settings of a production index to verify the distance metric. ## Benefits - Stop writing curl commands. Use search to run k-NN queries instantly through your chat interface. - Faster index prototyping. Create new vector structures with create_index without leaving your agent. - Easier debugging. Use get_index to check your mappings and settings in seconds rather than hunting through a dashboard. - Direct data management. Use index_document to add embeddings and delete_document to clean up your vector space. - Better visibility. Use list_indexes to see the health and document counts of your entire cluster at a glance. ## How It Works The bottom line is you get a conversational interface for your entire OpenSearch vector store. 1. Subscribe to the Connector and provide your OpenSearch host, username, and password. 2. Connect your AI client to the Vinkius marketplace. 3. Ask your agent to create indexes or search for similar embeddings. ## Frequently Asked Questions **Can the OpenSearch Vector MCP help me build a RAG system?** Yes, it is designed specifically for RAG. It allows your agent to query your vector store and retrieve relevant context documents for your generation pipeline. **How do I use OpenSearch Vector MCP to manage my embeddings?** You can manage them through natural conversation. Ask your agent to add new documents with metadata or delete specific entries from your vector space. **Does the OpenSearch Vector MCP support cosine similarity?** Yes, the Connector includes tools to create and manage indexes optimized specifically for cosine similarity and other k-NN metrics. **Can I use OpenSearch Vector MCP to check my cluster health?** You can. It provides a way to list all indexes and see their health status and document counts without opening a dashboard. **How does OpenSearch Vector MCP handle document deletion?** It allows you to remove specific vector documents from your OpenSearch cluster by providing the unique document ID to your agent. **Is OpenSearch Vector MCP good for ML engineers?** It is a great tool for ML engineers because it removes the need to write manual curl commands for testing similarity queries against production embeddings. **What vector dimensions does it support?** Any dimension supported by OpenSearch k-NN. Common values: 384 (MiniLM), 768 (BERT/all-mpnet), 1536 (OpenAI text-embedding-ada-002), 3072 (text-embedding-3-large). When creating an index, specify the exact dimension and the agent provisions the mapping automatically. **Can I delete an entire index or just individual documents?** Currently, the agent supports deleting individual documents by ID from an index. Full index deletion is not exposed through this integration to prevent accidental data loss. If you need to drop an index, use the OpenSearch Dashboards or direct API calls. **Does this work with Amazon OpenSearch Service (managed)?** Yes. Provide the Amazon OpenSearch Service endpoint as the host (e.g., https://search-xxx.us-east-1.es.amazonaws.com) along with the master username and password. The integration uses standard REST APIs that work identically on managed and self-hosted clusters.