MCP Servers

模型上下文协议服务器、框架、SDK 和模板的综合目录。

Z
Zed MCP Lancedb

MCP server by hansdoebel

创建于 2/24/2026
更新于 about 12 hours ago
Repository documentation and setup instructions

LanceDB MCP Server for Zed

A Zed extension that integrates LanceDB into Zed's Agent Panel via the Model Context Protocol (MCP). LanceDB is an embedded vector database (like SQLite for vectors) — no external server required.

Prerequisites

Download a GGUF embedding model and start the server:

llama-server -m your-model.gguf --embedding --port 8080

Installation

Install from the Zed Extensions panel: search for LanceDB MCP Server.

Configuration

Add to your Zed settings (~/.config/zed/settings.json):

{
  "context_servers": {
    "mcp-server-lancedb": {
      "settings": {
        "lancedb_uri": "~/lancedb"
      }
    }
  }
}

All settings are optional — sensible defaults are provided.

Settings

| Setting | Default | Description | |---------|---------|-------------| | lancedb_uri | ~/lancedb | Database path. Supports local paths, cloud (db://host:port), S3 (s3://...) | | lancedb_api_key | — | API key for LanceDB Cloud (only when URI starts with db://) | | table_name | lancedb-mcp-table | Default table name | | embedding_url | http://localhost:8080 | OpenAI-compatible embedding endpoint |

Tools

| Tool | Description | |------|-------------| | lancedb-list-tables | List all tables in the database | | lancedb-table-details | Get table metadata (row count, schema) | | lancedb-create-table | Create a new table with initial documents | | lancedb-delete-table | Drop a table | | lancedb-ingest | Ingest documents with auto-embedding (creates table if needed) | | lancedb-query | Search via vector, full-text, or hybrid mode | | lancedb-delete-rows | Delete rows matching a SQL WHERE filter | | lancedb-count-rows | Count rows with optional filter | | lancedb-fts-create-index | Create a full-text search index |

Query types

The lancedb-query tool supports three search modes via query_type:

  • vector (default) — semantic similarity search using embeddings
  • fts — full-text keyword search (requires index via lancedb-fts-create-index)
  • hybrid — combines vector + full-text results

Development

Build the WASM extension

cargo build --release --target wasm32-wasip1

Install as dev extension

Zed Command Palette → zed: install dev extension → select this directory.

Run tests

cd server && bun install && bun test

License

MIT

快速设置
此服务器的安装指南

安装包 (如果需要)

npx @modelcontextprotocol/server-zed-mcp-lancedb

Cursor 配置 (mcp.json)

{ "mcpServers": { "hansdoebel-zed-mcp-lancedb": { "command": "npx", "args": [ "hansdoebel-zed-mcp-lancedb" ] } } }