MCP Database Bridge
buvaneswaraneb/mcp-database-bridge
Consulta segura em SQLite pelo Claude com proteção contra operações destrutivas
Ver no GitHubServidor MCP leve que conecta Claude Desktop a bancos de dados SQLite locais com segurança. Permite análise de dados orientada por IA sem risco de operações destrutivas, com proteção via regex e modo somente leitura.
Instalação
npx convez-ai buvaneswaraneb/mcp-database-bridge1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Análise de dados com Claude sem expor bancos a alterações
- Exploração de schemas e estrutura de dados por IA
- Debugging de queries complexas com explicação automática
- Integração de Claude Desktop com aplicações que usam SQLite
Veredito

Ferramenta bem projetada para análise segura de dados com Claude. Proteção robusta contra operações destrutivas, setup automatizado para macOS/Linux/Windows, e suporte a introspection de schemas. Recomendado para quem precisa dar ao Claude acesso a bases SQLite sem riscos de modificação.
README
A tradução para o português deste README ainda está na fila.
Demo Video
Click the preview below to watch the MCP Database Bridge project demonstration on YouTube.
✨ Features
- 🛡️ Read-Only Safeties: Strict regex filtering intercepts and rejects destructive operations like
INSERT,UPDATE,DROP, andALTER. The AI can look, but it can't touch. - 🔍 Introspection: AI can autonomously list tables and read schemas (
PRAGMA table_info) directly to understand your data structure before writing queries. - 🚀 Claude Desktop Ready: Comes with one-click automated setup scripts for both macOS/Linux and Windows that instantly wire it up to your Claude Desktop config.
- 📊 Query Analysis: Includes
explain_querycapabilities to help AI debug complex data retrieval.
Quick Start
The setup scripts create a Python virtual environment, install dependencies, and register database-mcp in Claude Desktop.
Prerequisites
- Python 3.11 or newer
- Git
- Claude Desktop
1. Clone the Repository
git clone https://github.com/buvaneswaraneb/mcp-database-bridge.git
cd mcp-database-bridge
2. Run the Setup Script
Choose the instructions for your operating system.
Run:
chmod +x setup.sh
./setup.sh
After setup completes:
- Completely quit Claude Desktop.
- On macOS, press
Cmd + Q. - On Linux, quit Claude from the application menu.
- On macOS, press
- Reopen Claude Desktop.
- Open a new chat and confirm that
database-mcpappears in the available tools.
Run setup.bat from Command Prompt:
setup.bat
You can also double-click setup.bat from File Explorer.
Completely Restart Claude Desktop on Windows
Closing the Claude window may leave it running in the background. Fully stop it before reopening:
- Press
Ctrl + Shift + Escto open Task Manager. - Select the Processes tab.
- Find Claude under Apps or Background processes.
- Select each Claude process and click End task.
- Wait a few seconds, then reopen Claude Desktop.
If Claude does not appear under Processes, open the Details tab and end any Claude.exe processes.
3. Verify the Connection
In Claude Desktop:
- Open a new chat.
- Open the tools or integrations menu.
- Confirm that
database-mcpis connected. - Ask:
What databases are available?
If the server is not listed, completely stop Claude Desktop again and reopen it.
🛠️ Manual Setup (Claude Code / Custom Agents)
If you prefer to configure things manually or use Claude Code in your terminal:
-
Create Virtual Environment & Install Dependencies:
python3 -m venv .venv source .venv/bin/activate pip install -r requirements.txt cp .env.example .env -
Add to Claude Code:
claude mcp add database-mcp .venv/bin/python src/server.py
🧪 Running Tests
Ensure everything is working correctly by running the comprehensive test suite:
pytest tests/ -v
🏗️ Architecture & Internals
graph TD
A[🤖 Claude Desktop / AI Agents] <-->|stdio JSON-RPC| B{MCP Bridge Router}
subgraph Database MCP Server
B -->|tools/call: list_tables| C[📋 list_tables]
B -->|tools/call: get_schema| D[🔍 get_schema]
B -->|tools/call: explain_query| E[📊 explain_query]
B -->|tools/call: run_select| F[▶️ run_select]
F --> G{🛡️ Read-Only Safeties}
G -. Block UPDATE/DROP/INSERT .-> H[❌ Reject]
G -- Allow SELECT --> I[✅ Execute]
end
C & D & E & I --> J[(🗄️ SQLite Database)]
style A fill:#4B32C3,stroke:#fff,stroke-width:2px,color:#fff
style B fill:#2D3748,stroke:#4B32C3,stroke-width:2px,color:#fff
style G fill:#9B2C2C,stroke:#FC8181,stroke-width:2px,color:#fff
style J fill:#276749,stroke:#68D391,stroke-width:2px,color:#fff
style H fill:#E53E3E,stroke:#fff,stroke-width:1px,color:#fff
style C fill:#2A4365,stroke:#63B3ED,color:#fff
style D fill:#2A4365,stroke:#63B3ED,color:#fff
style E fill:#2A4365,stroke:#63B3ED,color:#fff
style F fill:#2A4365,stroke:#63B3ED,color:#fff
style I fill:#38A169,stroke:#fff,color:#fff
Curious how it works under the hood?
- Read docs/structure.md for a detailed walkthrough of the file architecture and the JSON-RPC execution flow.
- Read docs/ai_usage_note.md for notes on how AI was leveraged to build this project.