# Milvus MCP for AI Agents AI Agent Connect > Milvus (Open-Source Vector Database) MCP lets you manage vector storage, perform ANN searches, query scalar entities, and audit collections. Connect your database to any AI client to perform high-performance vector searches and monitor index health using natural language commands. ## Overview - **Category:** loved-by-devs - **Price:** Free - **Endpoint:** https://edge.vinkius.com/vk_preview_xJCavC8CNaxFuBWU15XnlYyQ3g4Cn853dcyYbM82/ai-agent-connect - **Tags:** vector-search, embeddings, similarity-search, ai-infrastructure, data-retrieval ## Description The Milvus MCP lets you connect your vector instance to your AI agent so you can manage high-performance search and embedding storage through natural conversation. Instead of jumping between your terminal and a web dashboard to check your indices or fetch specific data points, you can just ask your agent to do the heavy lifting. You get a direct line into your vector database where you can audit collection schemas, check real-time memory usage, and run complex searches that combine semantic similarity with structured filters like tags or dates. It takes the friction out of maintaining large-scale vector stores by letting you handle the data management tasks without writing manual scripts or boilerplate code. Whether you are debugging a specific record to see why it didn't rank correctly or monitoring the health of your production search, this Connector makes the data accessible in a way that fits into your existing workflow. It's one of the many tools you can find in the Vinkius catalog to keep your infrastructure manageable and your focus on the actual logic of your application. By moving these checks into your chat interface, you cut out the constant context switching that happens when you have to leave your workspace to verify a single dimension or row count. ## Tools ### list_collections See every vector collection currently tracked in your vector database. This helps you get a clear overview of your data structure at a glance. ### describe_collection View the explicit schema mapping and indexing definitions for a specific collection. This is useful for verifying that your dimensions and keys are set up correctly. ### search_vectors Perform nearest neighbor searches by providing an array of embedding vectors. This lets you find the most relevant matches for your user queries instantly. ### query_entities Retrieve specific entities using scalar expressions like tags or dates. This allows you to filter your search results by structured metadata. ### get_entities Extract unique vector items that match a specific primary key. You can use this to audit exact data points without relying on semantic similarity. ### get_collection_stats Get real-time metrics for row counts and physical memory usage. You can use this to monitor the health and scale of your production indices. ### delete_entities Irreversibly delete specific vector records using their primary identifiers. This makes it easy to keep your search index clean and optimized. ## Prompt Examples **Prompt:** ``` List all my vector collections. ``` **Response:** ``` I've retrieved the following collections from your Milvus instance: | Collection Name | Dimensions | Index Type | | :--- | :--- | :--- | | `image_embeddings` | 512 | IVF_FLAT | | `text_knowledge_base` | 1536 | HNSW | | `user_profiles` | 768 | IVF_FLAT | Which one would you like to inspect further? ``` **Prompt:** ``` Check the memory usage for 'user_profiles'. ``` **Response:** ``` Retrieving stats for `user_profiles`... **Collection Stats:** - **Entity Count:** 1,250,000 - **Memory Usage:** ~2.4 GB - **Status:** All indices are loaded and healthy. No anomalous data distribution detected. ``` **Prompt:** ``` Search for similar items to this vector: [0.1, -0.2, 0.5, ...] ``` **Response:** ``` Executing ANN search... I've identified the 5 nearest neighbors. The top match has a similarity score of **0.94** and maps to entity ID `12345`. Other results include: - ID `67890` (Score: 0.89) - ID `11223` (Score: 0.82) Would you like the full scalar data for these entities? ``` ## Capabilities ### Run similarity searches Get the closest matching vectors from your collections instantly. ### Filter by metadata Narrow down your search results using specific tags, IDs, or dates. ### Audit collection schemas See the exact structure and dimensions of your vector data. ### Monitor storage health Check how much memory your indices are using in real-time. ### Fetch specific records Grab individual items by their primary keys for precise data auditing. ### Clean up data Remove specific vector entries to keep your search index optimized. ## Use Cases ### Debugging Embedding Dimensions An ML engineer isn't sure if their new model's embeddings match the database. They ask the agent to describe the collection and check the dimensions. ### Production Monitoring A search architect needs to see if a recent data load spiked memory usage. They ask the agent to show the stats for the 'image_embeddings' collection. ### Hybrid Search Testing A developer wants to see results for 'blue shoes' but only for the 'summer' tag. They ask the agent to perform a search using search_vectors and query_entities. ### Data Auditing A data steward needs to find a specific user's vector to ensure privacy compliance. They ask the agent to fetch the item by its primary key. ## Benefits - Stop writing boilerplate: Use search_vectors to get semantic results without writing custom SDK code every time you need to test a query. - Faster debugging: Use describe_collection and get_collection_stats to quickly verify your schema and memory footprint in a chat window. - Precise filtering: Combine vector similarity with query_entities to filter by metadata like dates or IDs for much more accurate results. - Easier maintenance: Use delete_entities to prune your database and keep your search index clean and performant without manual SQL-like commands. - Better oversight: Get a full bird's-eye view of your data with list_collections to manage multiple environments from a single interface. ## How It Works The bottom line is you get a natural language interface for your vector database management and search. 1. Subscribe to the Milvus MCP and provide your Base URL and API Key or Zilliz Cloud Token. 2. Connect the Connector to your preferred client like Claude, Cursor, or Windsurf. 3. Ask your agent to perform searches, check stats, or manage your vector collections. ## Frequently Asked Questions **Can I use the Milvus MCP to check my database health?** Yes, you can use it to see real-time stats like row counts and memory usage. This helps you monitor the health and scale of your production indices without manual checks. **How does Milvus MCP handle metadata filters?** It allows your agent to combine vector searches with specific scalar filters like tags, IDs, or dates. This makes it easy to perform hybrid searches using natural language. **Can I delete specific records with this?** Yes, you can remove vector items by their primary keys to keep your index clean and optimized. This is much faster than deleting items through a web UI. **Is this good for checking my schema?** Definitely. You can get the full schema mapping and index definitions for any collection. It's a great way to verify that your dimensions and keys are set up correctly. **Can I use this with Zilliz Cloud?** Yes, the Connector supports both standard Milvus and Zilliz Cloud tokens, making it compatible with your existing cloud infrastructure. **How do I perform an ANN search through my agent?** Use the `search_vectors` tool by providing the collection name and a JSON float array matching the collection's dimensions. Your agent will perform an Approximate Nearest Neighbor search and return the most semantically relevant entities. **Can I filter results using structured fields instead of just vectors?** Yes. Use the `query_entities` tool with a Milvus-style filter expression. This allows you to retrieve entities based on primary keys, tags, or other scalar fields without necessarily performing a vector similarity search. **How do I check the schema and dimension requirements for a Milvus collection?** The `describe_collection` tool retrieves the complete schema mapping. Your agent will report the required vector dimensions, index types, and primary key names, helping you ensure your search queries are compatible with the database logic.