# googleapis/genai-toolbox [Health: Active]

**Category:** 🗄️ Databases  
**Repository:** https://github.com/googleapis/genai-toolbox  
**GitHub Stars:** 16741  
**Views:** 6  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/googleapis-genai-toolbox

## Description
Open source MCP server specializing in easy, fast, and secure tools for Databases.

## Claude Desktop Quick Installation
No confirmed install details yet, so no config is published. Follow the setup steps in the repository above.

## Documentation

## What googleapis/genai-toolbox MCP server does

googleapis/genai-toolbox provides database-facing MCP tools for agents, developer tools, and applications. It can expose generic operations for supported databases, including table discovery and SQL execution, without requiring each client integration to implement database-specific logic.

The server also acts as a framework for defining application-specific tools. Developers can describe approved actions, inputs, data sources, and SQL statements rather than giving an agent unrestricted access to a database. This makes it suitable for assistants that need narrowly scoped queries, structured access patterns, semantic search, or natural-language-to-SQL workflows.

Supported systems include Google Cloud services such as AlloyDB, BigQuery, Cloud SQL, Spanner, Firestore, and Knowledge Catalog. Other listed integrations include PostgreSQL, MySQL, MariaDB, SQL Server, Oracle, MongoDB, Redis, Elasticsearch, CockroachDB, ClickHouse, Couchbase, Neo4j, Snowflake, and Trino.

## How it works

A client starts the server through MCP configuration, commonly using the `@toolbox-sdk/server` package with a `--prebuilt=<database>` argument and `--stdio`. The selected prebuilt configuration exposes standard tools for that database. A narrower toolset can be selected with a path such as `--prebuilt=postgres/data`.

For custom deployments, a `tools.yaml` file describes the available resources. Sources contain connection details, tools define agent actions and their parameters, and toolsets group tools for different agents or applications. Prompts can also be declared for LLM interactions. The server supports connection pooling, IAM-based authentication, and OpenTelemetry metrics and tracing.

## Setup and configuration

Add the server to the MCP client's configuration file, such as `mcp.json` or `claude_desktop_config.json`. A prebuilt PostgreSQL entry can invoke:

```json
{
  "command": "npx",
  "args": ["-y", "@toolbox-sdk/server", "--prebuilt=postgres", "--stdio"]
}
```

Set the connection environment variables associated with the selected prebuilt database. For custom tools, point the server at a YAML configuration with `--config tools.yaml`; multiple configuration files can be selected this way as well.

The project also supports other execution methods, including Docker and binaries. Client connections can use MCP directly, while application developers can integrate through the published Python, JavaScript/TypeScript, Go, and Java SDKs.

## Tools and capabilities

- Prebuilt generic tools for database access
- Table listing and SQL execution
- Database schema exploration and database-aware code generation
- Custom tools backed by predefined SQL or other configured logic
- Structured queries, semantic search, and NL2SQL tool patterns
- Toolsets for loading different groups of tools
- Prompt definitions for LLM interactions
- Connection pooling, IAM authentication, and OpenTelemetry observability

_Full upstream README: https://allmcps.com/mcp/googleapis-genai-toolbox/readme_