Model Context Protocol Servers for Milvus
The Model Context Protocol (MCP) is an open protocol that enables seamless integration between LLM applications and external data sources and tools. Whether you're building an AI-powered IDE, enhancing a chat interface, or creating custom AI workflows, MCP provides a standardized way to connect LLMs with the context they need.
This repository contains a MCP server that provides access to Milvus vector database functionality.
Before using this MCP server, ensure you have:
The recommended way to use this MCP server is to run it directly with uv without installation. This is how both Claude Desktop and Cursor are configured to use it in the examples below.
If you want to clone the repository:
git clone https://github.com/zilliztech/mcp-server-milvus.git
cd mcp-server-milvus
Then you can run the server directly:
uv run src/mcp_server_milvus/server.py --milvus-uri http://localhost:19530
Alternatively you can change the .env file in the src/mcp_server_milvus/ directory to set the environment variables and run the server with the following command:
uv run src/mcp_server_milvus/server.py
This MCP server can be used with various LLM applications that support the Model Context Protocol:
Install Claude Desktop from https://claude.ai/download
Open your Claude Desktop configuration:
~/Library/Application Support/Claude/claude_desktop_config.jsonAdd the following configuration:
{
"mcpServers": {
"milvus": {
"command": "/PATH/TO/uv",
"args": [
"--directory",
"/path/to/mcp-server-milvus/src/mcp_server_milvus",
"run",
"server.py",
"--milvus-uri",
"http://localhost:19530"
]
}
}
}
Cursor also supports MCP tools. You can add the Milvus MCP server to Cursor in two ways:
Go to Cursor Settings > Features > MCP
Click on the + Add New MCP Server button
Fill out the form:
stdio (since you're running a command)milvus/PATH/TO/uv --directory /path/to/mcp-server-milvus/src/mcp_server_milvus run server.py --milvus-uri http://127.0.0.1:19530⚠️ Note: Use
127.0.0.1instead oflocalhostto avoid potential DNS resolution issues.
Create a .cursor/mcp.json file in your project root:
Create the .cursor directory in your project root:
mkdir -p /path/to/your/project/.cursor
Create a mcp.json file with the following content:
{
"mcpServers": {
"milvus": {
"command": "/PATH/TO/uv",
"args": [
"--directory",
"/path/to/mcp-server-milvus/src/mcp_server_milvus",
"run",
"server.py",
"--milvus-uri",
"http://127.0.0.1:19530"
]
}
}
}
Restart Cursor or reload the window
After adding the server, you may need to press the refresh button in the MCP settings to populate the tool list. The Agent will automatically use the Milvus tools when relevant to your queries.
To verify that Cursor has successfully integrated with your Milvus MCP server:
The server provides the following tools:
milvus_text_search: Search for documents using full text search
collection_name: Name of collection to searchquery_text: Text to search forlimit: Maximum results (default: 5)output_fields: Fields to include in resultsdrop_ratio: Proportion of low-frequency terms to ignore (0.0-1.0)milvus_vector_search: Perform vector similarity search on a collection
collection_name: Name of collection to searchvector: Query vectorvector_field: Field containing vectors to search (default: "vector")limit: Maximum results (default: 5)output_fields: Fields to include in resultsmetric_type: Distance metric (COSINE, L2, IP) (default: "COSINE")milvus_query: Query collection using filter expressions
collection_name: Name of collection to queryfilter_expr: Filter expression (e.g. 'age > 20')output_fields: Fields to include in resultslimit: Maximum results (default: 10)milvus_list_collections: List all collections in the database
milvus_create_collection: Create a new collection with specified schema
collection_name: Name for the new collectioncollection_schema: Collection schema definitionindex_params: Optional index parametersmilvus_load_collection: Load a collection into memory for search and query
collection_name: Name of collection to loadreplica_number: Number of replicas (default: 1)milvus_release_collection: Release a collection from memory
collection_name: Name of collection to releasemilvus_insert_data: Insert data into a collection
collection_name: Name of collectiondata: Dictionary mapping field names to lists of valuesmilvus_delete_entities: Delete entities from a collection based on filter expression
collection_name: Name of collectionfilter_expr: Filter expression to select entities to deleteMILVUS_URI: Milvus server URI (can be set instead of --milvus-uri)MILVUS_TOKEN: Optional authentication tokenMILVUS_DB: Database name (defaults to "default")To run the server directly:
uv run server.py --milvus-uri http://localhost:19530
What are the collections I have in my Milvus DB?
Claude will then use MCP to check this information on your Milvus DB.
I'll check what collections are available in your Milvus database.
Here are the collections in your Milvus database:
1. rag_demo
2. test
3. chat_messages
4. text_collection
5. image_collection
6. customized_setup
7. streaming_rag_demo
Find documents in my text_collection that mention "machine learning"
Claude will use the full-text search capabilities of Milvus to find relevant documents:
I'll search for documents about machine learning in your text_collection.
> View result from milvus-text-search from milvus (local)
Here are the documents I found that mention machine learning:
[Results will appear here based on your actual data]
In Cursor, you can ask:
Create a new collection called 'articles' in Milvus with fields for title (string), content (string), and a vector field (128 dimensions)
Cursor will use the MCP server to execute this operation:
I'll create a new collection called 'articles' with the specified fields.
Collection 'articles' has been created successfully with the following schema:
- title: string
- content: string
- vector: float vector[128]
If you see errors like "Failed to connect to Milvus server":
docker ps (if using Docker)127.0.0.1 instead of localhost in the URIIf you see authentication errors:
MILVUS_TOKEN is correctIf the MCP tools don't appear in Claude Desktop or Cursor:
If you continue to experience issues: