Milvus MCP Server

@danchev

代码仓库

基本信息

分类

数据库

传输方式

stdio

作者

danchev

提交者

Dobri Danchev

关于 Milvus MCP Server

A Model Context Protocol (MCP) server for agentic retrieval and semantic search over unstructured and structured data using Milvus, a high-performance vector database. This server enables large language model (LLM) applications to efficiently index, store, and retrieve vector emb

#milvus #document-search #vector-database#text-search#vector-search#similarity-search#embeddings #ai-search #machine-learning#high-dimensional-search#data-retrieval

Milvus MCP Server 配置

使用下面的配置,将此服务器添加到你的 MCP 客户端。

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

Milvus MCP Server 工具

工具列表暂未提供

当前收录暂时没有可展示的工具列表，不代表服务没有工具。请查阅项目文档了解支持的能力。

点击上方按钮,在临时沙箱中运行此服务器并实时获取工具列表。

概览

What is Milvus?

Milvus is a Model Context Protocol (MCP) server that enables agentic retrieval and semantic search across structured and unstructured data. It is built on top of Milvus, a high-performance vector database optimized for similarity search at scale.

How to use Milvus?

Milvus provides a set of tools accessible through the MCP interface. Users invoke tools by name with their parameters, such as milvus_text_search for full-text search or milvus_vector_search for vector similarity search. No installation or configuration instructions are provided in the README.

Key features of Milvus

Full-text search on collections via milvus_text_search
Vector similarity search with distance metrics (COSINE, L2, IP)
Filter-based querying using expressions like milvus_query
Collection management: list, create, load, release, and inspect
Data insert and delete operations on collections

Use cases of Milvus

Searching documents using full-text queries across a Milvus collection
Finding similar items with vector similarity search (e.g., semantic search)
Filtering and retrieving entities based on structured conditions
Managing collection schemas and loading/unloading for performance

常见问题

What tools does Milvus provide?

Milvus provides tools for search and query operations (`milvus_text_search`, `milvus_vector_search`, `milvus_query`), collection management (`milvus_list_collections`, `milvus_create_collection`, `milvus_load_collection`, `milvus_release_collection`, `milvus_get_collection_info`), and data operations (`milvus_insert_data`, `milvus_delete_entities`).

What are the parameters for `milvus_vector_search`?

`milvus_vector_search` accepts `collection_name`, `vector` (the query vector), `vector_field` (default: "vector"), `limit` (default: 5), `output_fields`, and `metric_type` (default: "COSINE").

How do I delete entities from a collection?

Use the `milvus_delete_entities` tool with `collection_name` and a `filter_expr` (e.g., 'age > 20') to select which entities to delete.

What is the default result limit for text search?

The default limit for `milvus_text_search` is 5 results.

Does the server require a running Milvus instance?

The README does not specify runtime dependencies or whether an external Milvus instance must be running; it only states the server is built on top of Milvus.