seekdb MCP server

seekdb MCP server implements Anthropic's MCP (Model Context Protocol), enabling AI assistants to interact with a seekdb database and translate natural language into database operations. You describe what you want in plain language, and the assistant can create collections, insert data, run complex queries, and even help you build a complete knowledge base workflow.

This topic introduces the capabilities and usage of seekdb MCP server. For more information about seekdb, see seekdb database.

Architecture and features

Core capabilities

seekdb MCP server exposes database tools directly to AI assistants.

Category	Tools	Description
Vector collection management	`create_collection`, `query_collection`, `add_data_to_collection`, etc.	Create vector collections, run semantic search, and manage documents
Advanced search	`full_text_search`, `hybrid_search`	Full-text search and hybrid search (BM25 + vectors)
AI functions	`ai_complete`, `ai_rerank`, `create_ai_model`, etc.	Call LLMs for text generation and rerank search results
AI memory system	`seekdb_memory_query`, `seekdb_memory_insert`, etc.	Persistent memory across sessions so the assistant can "remember" you
Data import/export	`import_csv_file_to_seekdb`, `export_csv_file_from_seekdb`	Convert CSV files to and from database tables / vector collections

Vector collection management

Tool	Description
`create_collection`	Create a vector collection
`list_collections`	List all collections
`has_collection`	Check whether a collection exists
`peek_collection`	Preview documents in a collection
`add_data_to_collection`	Add documents (embeddings are generated automatically)
`update_collection`	Update documents
`delete_documents`	Delete documents
`query_collection`	Vector similarity search
`delete_collection`	Delete a collection

Advanced search

Tool	Description
`full_text_search`	Full-text search (keyword-based)
`hybrid_search`	Hybrid search (full-text + vector search)

AI model tools

Tool	Description
`create_ai_model`	Register an AI model (embedding, text generation, or reranking)
`create_ai_model_endpoint`	Create an endpoint that connects a model to an API service
`drop_ai_model`	Remove a registered AI model
`drop_ai_model_endpoint`	Remove an AI model endpoint
`ai_complete`	Generate text using an LLM
`ai_rerank`	Rerank documents by relevance using an AI model
`get_registered_ai_models`	List all registered AI models
`get_ai_model_endpoints`	List all AI model endpoints

AI memory system

seekdb MCP server provides an AI memory layer that allows assistants to remember information across sessions.

Tool	Description
`seekdb_memory_query`	Semantic search over memory
`seekdb_memory_insert`	Store new memories
`seekdb_memory_update`	Update memories
`seekdb_memory_delete`	Delete memories

Common use cases:

Remember your preferred stack (for example, "I prefer Python").
Remember project context (for example, "This project uses FastAPI").
Remember personal preferences (for example, "I prefer clean code style").

Data import/export

Tool	Description
`import_csv_file_to_seekdb`	Import a CSV file into a vector collection
`export_csv_file_from_seekdb`	Export data from a vector collection to a CSV file

SQL operations

Tool	Description
`execute_sql`	Execute an SQL query
`get_current_time`	Get the current time from the database

Installation and configuration

Install the seekdb database

Before you use seekdb MCP Server, you need a running seekdb database. seekdb supports two deployment modes. For details, see Deploy seekdb.

Embedded mode (zero configuration; Linux only)

If you use embedded mode, you do not need to install seekdb separately. When you start seekdb MCP server, it initializes a local embedded database automatically.

This is suitable for: personal learning, rapid prototyping, and edge devices.

tip
Embedded mode is currently supported on Linux only. If you are on macOS or Windows, use seekdb client/server mode and deploy seekdb first (Docker is recommended).
Client/server mode (recommended for production)

For details, see Deploy seekdb using yum install and Deploy seekdb in a container environment.

Install seekdb MCP server

Install uv

curl -LsSf https://astral.sh/uv/install.sh | sh

Configure your AI tool

Stdio transport

Using Cursor as an example: open Settings → Tools & MCP → New MCP Server, then choose the configuration based on your OS.

Linux (embedded mode):

{
"mcpServers": {
    "seekdb": {
    "command": "uvx",
    "args": ["seekdb-mcp-server"]
    }
}
}

When the server starts, it automatically initializes a local embedded seekdb database.

macOS / Windows (client/server mode):

macOS and Windows do not support embedded mode. Deploy seekdb first (Docker is recommended), then configure the connection parameters:

{
"mcpServers": {
    "seekdb": {
    "command": "uvx",
    "args": ["seekdb-mcp-server"],
    "env": {
        "SEEKDB_HOST": "127.0.0.1",
        "SEEKDB_PORT": "2881",
        "SEEKDB_USER": "root",
        "SEEKDB_PASSWORD": "",
        "SEEKDB_DATABASE": "test"
    }
    }
}
}

Parameter reference:

Parameter	Description	Default
`SEEKDB_HOST`	seekdb server host	`127.0.0.1`
`SEEKDB_PORT`	seekdb server port	`2881`
`SEEKDB_USER`	Database username	`root`
`SEEKDB_PASSWORD`	Database password	(empty)
`SEEKDB_DATABASE`	Database name	`test`

SSE transport

Linux (embedded mode)

Start the SSE server:

uvx seekdb-mcp-server --transport sse --port 6000

macOS / Windows (client/server mode)

Configure environment variables, then start the server:

export SEEKDB_HOST=127.0.0.1
export SEEKDB_PORT=2881
export SEEKDB_USER=root
export SEEKDB_PASSWORD=
export SEEKDB_DATABASE=test

Start the SSE server:

uvx seekdb-mcp-server --transport sse --port 6000

Configure the client:

{
"sse-seekdb": {
    "type": "sse",
    "url": "http://127.0.0.1:6000/sse"
}
}

Application scenarios

seekdb MCP server is suitable for a wide range of use cases.

Build a knowledge base

Store documents, notes, and code snippets in a vector collection and find relevant content with semantic search:

"Add this article about React Hooks to the knowledge base"
"Search for previously saved notes on state management"

Prototype a RAG app

Validate ideas for RAG (retrieval-augmented generation) applications without building a complex backend:

"Create a collection of product documents and import them"
"Answer user questions based on these documents"

Data analysis

Import CSV data and analyze it using natural language:

"Import this sales data CSV file"
"Analyze last month's sales trends"

Enhance AI coding assistants

Give AI coding assistants persistent memory:

"Remember: This project uses PostgreSQL as the primary database"
"Remember: I prefer TypeScript over JavaScript"

More operations

To try a hands-on workflow end to end, see Build a personal knowledge base with seekdb MCP and AI chat.

Architecture and features​

Core capabilities​

Vector collection management​

Advanced search​

AI model tools​

AI memory system​

Data import/export​

SQL operations​

Installation and configuration​

Install the seekdb database​

Install seekdb MCP server​

Install uv​

Configure your AI tool​

Stdio transport​

SSE transport​

Application scenarios​

Build a knowledge base​

Prototype a RAG app​

Data analysis​

Enhance AI coding assistants​

More operations​

Contents