Milvus MCP Server

Connect AI assistants to Milvus, the open-source vector database, for semantic search, vector storage, and hybrid query operations through the Model Context Protocol.

May 4, 2026
GitHub starsPyPI version

Overview

The Milvus MCP Server enables AI assistants to interact with Milvus, the high-performance open-source vector database built for scalable similarity search and AI applications. Through the Model Context Protocol, AI models can create and manage vector collections, perform semantic search, insert data, and manage indexes — making Milvus's vector storage capabilities available directly through natural language.

Official Server:

Developed by Zilliz, the team behind Milvus

Key Features

🔍

Vector Search

Perform high-dimensional vector similarity search across collections with configurable top-k results.

🗂️

Collection Management

Create, describe, list, and drop collections with custom schemas and vector dimensions.

📥

Data Ingestion

Insert vectors and associated metadata into collections for retrieval and search.

Index Management

Create and manage vector indexes to optimize search performance at scale.

🔗

Hybrid Search

Combine vector similarity with attribute filtering for precise, context-aware results.

☁️

Cloud & Self-Hosted

Connect to Milvus Lite, standalone deployments, or fully managed Zilliz Cloud clusters.

Available Tools

Quick Reference

ToolPurposeCategory
create_collectionCreate a new collection with schema definitionSchema
describe_collectionGet collection details and statisticsDiscovery
list_collectionsList all collections in the databaseDiscovery
drop_collectionDelete a collection and its dataSchema
insert_dataInsert vectors with metadata into a collectionWrite
vector_searchSearch for similar vectors with optional filteringSearch
create_indexBuild an index for faster vector searchIndexing

Detailed Usage

create_collection

Create a new collection with a defined schema including vector field dimension and optional metadata fields.

use_mcp_tool({
  server_name: "milvus",
  tool_name: "create_collection",
  arguments: {
    collection_name: "documents",
    dimension: 1536,
    description: "Document embeddings collection"
  }
});
vector_search

Search for the most similar vectors in a collection, with optional metadata filtering.

use_mcp_tool({
  server_name: "milvus",
  tool_name: "vector_search",
  arguments: {
    collection_name: "documents",
    vector: [0.1, 0.2, ...],
    top_k: 10,
    filter: "category == 'research'"
  }
});
insert_data

Insert vectors with associated metadata into a collection for future search and retrieval.

use_mcp_tool({
  server_name: "milvus",
  tool_name: "insert_data",
  arguments: {
    collection_name: "documents",
    vectors: [[0.1, 0.2, ...], [0.3, 0.4, ...]],
    metadata: [
      { text: "First document", source: "pdf" },
      { text: "Second document", source: "web" }
    ]
  }
});
create_index

Create an index on a collection to accelerate vector similarity search.

use_mcp_tool({
  server_name: "milvus",
  tool_name: "create_index",
  arguments: {
    collection_name: "documents",
    index_type: "IVF_FLAT",
    metric_type: "IP",
    params: { nlist: 1024 }
  }
});

Installation

{
  "mcpServers": {
    "milvus": {
      "command": "uvx",
      "args": [
        "mcp-server-milvus"
      ],
      "env": {
        "MILVUS_URI": "http://localhost:19530"
      }
    }
  }
}

Connection Options:

Set MILVUS_URI to your Milvus server address. Use http://localhost:19530 for local Milvus, or your Zilliz Cloud endpoint for managed deployments. Add MILVUS_TOKEN for authentication when needed.

Common Use Cases

  • AI-Powered RAG Pipelines: Connect vector stores to LLMs for retrieval-augmented generation with semantic search
  • Document & Media Search: Index and query embeddings from documents, images, or audio for intelligent retrieval
  • Recommendation Systems: Build vector-based similarity search for product or content recommendations
  • Multi-Agent Knowledge Bases: Share a common vector store across multiple AI agents for consistent context

Sources