bytebase/dbhub
bytebase/dbhub has 3,792 GitHub stars and 331 forks. GitHot measured +6 stars in 24 hours and +39 in 7 days. Last checked 2026-11-05 18:43 UTC. Last push 2026-10-26.
注重 Token 使用量的数据库 MCP 服务器,支持 Postgres、MySQL、SQL Server、Oracle、MariaDB 和 SQLite。
实测星标增长
- 24小时
- +6
- 7天
- +39
- 15天
- 正在收集历史
- 1个月
- 正在收集历史
各周期内的星标净变化,周期相互重叠,并非每日新增。缺失的测量保持空白。星标代表关注度,不代表软件质量。
README.md
原文[!NOTE]
If you need an enterprise-level database MCP server with built-in guardrails like approval flow, access control, data masking, and audit logging beyond what DBHub offers, check out Bytebase.
+------------------+ +--------------+ +------------------+
| | | | | |
| Claude Desktop +--->+ +--->+ PostgreSQL |
| | | | | |
| Claude Code +--->+ +--->+ SQL Server |
| | | | | |
| Cursor +--->+ DBHub +--->+ Oracle |
| | | | | |
| VS Code +--->+ +--->+ SQLite |
| | | | | |
| Copilot CLI +--->+ +--->+ MySQL |
| | | | | |
| | | +--->+ MariaDB |
| | | | | |
+------------------+ +--------------+ +------------------+
MCP Clients MCP Server Databases
DBHub is a minimal MCP server: token-efficient, zero-dependency, and just two tools by default with opt-in extras. This lightweight gateway allows MCP-compatible clients to connect to and explore different databases:
- Minimal: Zero dependency, token efficient with a minimal set of MCP tools to maximize context window
- Multi-Database: PostgreSQL, MySQL, MariaDB, SQL Server, Oracle, and SQLite through a single interface
- Multi-Connection: Connect to multiple databases simultaneously with TOML configuration
- Guardrails: Read-only mode, row limiting, and query timeout to prevent runaway operations
- Secure Access: SSH tunneling and SSL/TLS encryption
DBHub is the official example in the Claude Code docs for connecting to PostgreSQL via MCP.
Token Efficiency
DBHub loads just 2 tools by default at 1.4k tokens — 13-14x fewer than alternatives — keeping the context window open for your actual work.
| MCP Server | Default Config | Default Tools |
|---|---|---|
| DBHub | 1.4k | 2 (execute_sql, search_objects) |
| MCP Toolbox | 19.0k | 28 |
| Supabase MCP | 19.3k | all |
Use Cases
- Local Development: Schema exploration, query validation, and data debugging with Claude Code, VS Code, Cursor, etc.
- Non-Technical Access: Expose curated, read-only views to non-technical staff via Claude Desktop, VS Code, Cursor, etc.
- Multi-Database Consolidation: Replace separate MCP servers for each database with a single DBHub process
- Production Troubleshooting: Read-only diagnostics with guardrails against runaway queries
Supported Databases
PostgreSQL, MySQL, SQL Server, MariaDB, Oracle, and SQLite.
MCP Tools
DBHub implements MCP tools for database operations:
- execute_sql: Execute SQL queries with transaction support and safety controls
- search_objects: Search and explore database schemas, tables, columns, indexes, and procedures with progressive disclosure
- explain_sql (opt-in): Show a query's execution plan without running it
- health_check (opt-in): Report connection pool state and buffer cache hit ratio
- Custom Tools: Define reusable, parameterized SQL operations in your
dbhub.tomlconfiguration file
Workbench
DBHub includes a built-in web interface for interacting with your database tools. It provides a visual way to execute queries, run custom tools, and view request traces without requiring an MCP client.

Installation
npx @bytebase/dbhub@latest --transport http --port 8080 --dsn "postgres://user:password@localhost:5432/dbname?sslmode=disable"
Also available as:
- Docker image
- MCP Bundle (one-click install, read-only)
- Claude Code plugin
See the Installation Guide for all options, Command-Line Options for parameters, and Multi-Database Configuration for connecting several databases at once.
Development
Requires Node.js >= 22.5.0 (DBHub uses the built-in node:sqlite module).
# Install dependencies
pnpm install
# Run in development mode
pnpm dev
# Build and run for production
pnpm build && pnpm start --transport stdio --dsn "postgres://user:password@localhost:5432/dbname"
Contributors
Star History
探索相关项目
Asabeneh/30-Days-Of-Python
30 Days of Python 编程挑战是一份循序渐进的指南,用于在 30 天内学习 Python 编程语言。这项挑战可能需要超过 100 天。请按自己的节奏进行。这些视频或许也有帮助:https://www.youtube.com/channel/UC7PNRuno1rzYPb1xLa4yktw
78,163 星标grafana/grafana
开放且可组合的可观测性与数据可视化平台。可对来自 Prometheus、Loki、Elasticsearch、InfluxDB、Postgres 等众多数据源的指标、日志和追踪进行可视化。
77,699 星标redis/redis
对于正在构建实时数据驱动应用的开发者而言,Redis 是首选、速度最快且功能最丰富的缓存、数据结构服务器和文档与向量查询引擎。
77,199 星标apache/superset
Apache Superset 是一个数据可视化与数据探索平台
75,413 星标