Spanner

Note

Currently in beta (pre-v1.0), and may see breaking changes until the first stable release (v1.0).

This repository packages MCP Toolbox's prebuilt spanner server as a plugin/extension to interact with Google Cloud Spanner instances. It can be used with various AI agents, including Antigravity, Claude Code and Codex, to manage your databases, execute queries, explore schemas, and troubleshoot issues using natural language prompts.

Important

We Want Your Feedback! Please share your thoughts with us by filling out our feedback form. Your input is invaluable and helps us improve the project for everyone.

Table of Contents

• Why Use Spanner?
• Prerequisites
• Getting Started
  • Configuration
  • Installation & Usage
    • Antigravity
    • Claude Code
    • Codex
• Installing via a compatible Agent Plugins client
• Usage Examples
• Available Tools
• Generating Skills Instead
• Additional Extensions
• Troubleshooting

Why Use Spanner?

• Seamless Workflow: Integrates seamlessly into your AI agent's environment. No need to constantly switch contexts for common database tasks.
• Natural Language Queries: Stop wrestling with complex commands. Explore schemas and query data by describing what you want in plain English.
• Full Lifecycle Control: Manage your Spanner databases, from exploring schemas to running queries.
• Code Generation: Accelerate development by asking your agent to generate data classes and other code snippets based on your table schemas.

Prerequisites

Before you begin, ensure you have the following:

• One of these AI agents installed
  • Antigravity
    • Antigravity CLI version v1.6.0 or higher
    • Antigravity 2.0 version v2.0.0 or higher.
  • Claude Code version v2.1.94 or higher.
  • Codex v0.117.0 or higher.
• Node.js — the MCP server runs via npx.
• A Google Cloud project with the Spanner API enabled.
• Ensure Application Default Credentials are available in your environment.
• IAM Permissions:
  • Cloud Spanner Database Reader (roles/spanner.databaseReader)
  • Cloud Spanner Database User (roles/spanner.databaseUser)

Getting Started

Configuration

Please keep these env vars handy during the installation process:

• SPANNER_PROJECT: The GCP project ID.
• SPANNER_INSTANCE: The Spanner instance ID.
• SPANNER_DATABASE: The Spanner database ID.
• SPANNER_DIALECT: (Optional) The SQL dialect of the Spanner Database: 'googlesql' or 'postgresql'. Defaults to "googlesql".

Note

• Ensure Application Default Credentials are available in your environment.

Installation & Usage

To start interacting with your database, install the extension for your preferred AI agent, then launch the agent and use natural language to ask questions or perform tasks.

For the latest version, check the releases page.

Antigravity

You can use either of these two agents for Antigravity:

• Antigravity CLI version v1.6.0 or higher
• Antigravity 2.0 version v2.0.0 or higher.

💡 Tip — Migrating from Gemini CLI?
If you previously installed this extension with gemini extensions install, you can convert it to an Antigravity plugin instead of reinstalling from scratch:

• On first launch of Antigravity CLI, accept the Migration Options prompt to automatically convert your installed Gemini CLI extensions to Antigravity plugins.
• Or, from your terminal, run:

  agy plugin import gemini

See Migrating from Gemini CLI for details on plugins, context files (GEMINI.md / AGENTS.md), and MCP server config differences.

Antigravity 2.0 (IDE)

1. Install the plugin:

Install the plugin directly from the remote GitHub repository:

agy plugin install https://github.com/gemini-cli-extensions/spanner

2. Set env vars: Set your environment vars as described in the configuration section.

(Tip: You can verify the MCP server is active by running the /mcp command in your active session.)

Antigravity CLI

You can install plugins directly from a remote GitHub repository.

1. Install the plugin:

agy plugin install https://github.com/gemini-cli-extensions/spanner

2. Set env vars: Set your environment vars as described in the configuration section.

Claude Code

1. Set env vars: In your terminal, set your environment vars as described in the configuration section.

2. Start the agent:

claude

3. Install the plugin:

/plugin install spanner@claude-plugins-official

_(Tip: Run /plugin list inside Claude Code to verify the plugin is active, or /reload-plugins if you just installed it.)

Codex

1. Install marketplace:

codex plugin marketplace add GoogleCloudPlatform/data-agent-kit

2. Install the plugin:

codex plugin add spanner@data-agent-kit

3. Set env vars: Enter your environment vars as described in the configuration section.

4. (Optional) Update the marketplace:

codex plugin marketplace upgrade data-agent-kit

Installing via a compatible Agent Plugins client

Installing via a compatible Agent Plugins client

This repository is a valid Agent Plugins (v1) plugin. Any Agent Plugins–compatible client can install it directly using its own built-in plugin command — no extra tooling required — by pointing at this repository:

https://github.com/gemini-cli-extensions/spanner

Beyond harnesses covered by the native install above, compatible clients include VS Code, Cursor, GitHub Copilot, and Kiro. See your agent's documentation for its exact install command.

Set env vars: Set your environment vars as described in the configuration section.

Usage Examples

Interact with Spanner using natural language:

• Explore Schemas and Data:
  • "Show me all tables in the 'orders' database."
  • "What are the columns in the 'products' table?"
  • "How many orders were placed in the last 30 days, and what were the top 5 most purchased items?"
• Generate Code:
  • "Generate a Python dataclass to represent the 'customers' table."

Available Tools

The tools come from MCP Toolbox's prebuilt spanner server, grouped into toolsets:

• data - Use these tools when you need to explore the database structure, discover schema objects like tables and graphs, and execute custom SQL queries to interact with your data.

For the full, up-to-date list, see the spanner prebuilt config in the MCP Toolbox repository.

Generating Skills Instead

The tool-backed skills this plugin used to ship were generated from the same prebuilt toolsets. If your agent lacks deferred tool loading, or you prefer skills, regenerate them with the script in this repository:

VERSION=<toolbox version> ./.github/scripts/generate_skills.sh

Use the toolbox version pinned in mcp.json. A single toolset, without the script:

npx @toolbox-sdk/server@<toolbox version> --prebuilt spanner skills-generate \
  --name "<skill name>" \
  --toolset "<toolset>" \
  --description "<what it is for>"

The generated scripts call the toolbox through npx, so no binary download is needed. See Generate Agent Skills in the MCP Toolbox repository.

Additional Extensions

Find additional extensions to support your entire software development lifecycle at github.com/gemini-cli-extensions.

Troubleshooting

Use the debug mode of your agent (e.g., gemini --debug) to enable debugging.

Common issues:

• "failed to find default credentials: google: could not find default credentials.": Ensure Application Default Credentials are available in your environment. See Set up Application Default Credentials for more information.
• "✖ Error during discovery for server: MCP error -32000: Connection closed": The database connection has not been established. Ensure your configuration is set via environment variables.
• "✖ MCP ERROR: Error: spawn npx ENOENT": Node.js is not installed, or npx is not on your PATH. Install Node.js, which provides npx.
• "cannot execute binary file": The Toolbox binary did not download correctly. Ensure the correct binary for your OS/Architecture has been downloaded. See Installing the server for more information.