Google Maps Platform agent skills ([GitHub](https://github.com/googlemaps/agent-skills)) are portable, self-contained modules of Google Maps Platform-specific knowledge, instructions, and workflows. They're designed to help AI assistants understand Google Maps Platform best practices and execute complex tasks with higher accuracy and lower token cost. ## Available skills Google Maps Platform provides a suite of skills for core products. Skills for additional products and features are coming soon! | Skill | Description | |---|---| | `google-maps-platform` | A collection of skills for architecting and implementing production-ready code using Google Maps Platform APIs and SDKs across Web, Android, iOS, and Web Services APIs for any of the following: map, place, address, geocoding, routing/ETA (including eco-friendly routing), nearby search, 3D/Street View/static map, marker clustering, custom styling, drawing, geofencing, heatmap, or environmental (air-quality/pollen/solar/weather). | > [!NOTE] > **Note:** For prototyping, use the public [Maps Demo Key](https://developers.google.com/maps/demo-key.md.txt)---no billing setup, no Cloud project required, and it covers a growing set of the most popular Google Maps Platform APIs. For production, the skill prompts you to create and restrict your own key. All non-trivial code is grounded in freshly retrieved docs via the Google Maps Platform Code Assist service (no reliance on training-data memory). ## Get started Google Maps Platform agent skills work with any AI assistant that supports skills, including Gemini, Cursor, Windsurf, Copilot, and Claude. ### Install Google Maps Platform agent skills In most cases, Google Maps Platform agent skills can be installed with a single command: ### Skills CLI For [most popular AI-assistive tools](https://www.skills.sh/agent), you can use the `skills` CLI to install Google Maps Platform agent skills. Run the following in the root directory of your project: ```bash npx skills add googlemaps/agent-skills ``` This method has been tested as a terminal command or a natural language prompt to the agent in: - **AI Studio** - **Antigravity** - **Claude Code** - **Replit** (alternate instructions: [Use the Project Editor UI](https://docs.replit.com/build/use-agent-skills#install-a-skill-in-your-project)) ### Gemini CLI This repository is configured as a [Gemini CLI extension](https://geminicli.com/extensions/about/). You can add it using the Gemini CLI: ```bash gemini extensions install https://github.com/googlemaps/agent-skills ``` ### Lovable Install the Google Maps Platform agent skills in Lovable [using the Skills settings UI](https://docs.lovable.dev/features/skills#import-from-github): 1. Visit `https://lovable.dev/settings/skills` 2. Under Workspace skills, click **Add** \> **Import from GitHub**. 3. Paste the URL for the `google-maps-platform` main skill: https://github.com/googlemaps/agent-skills/tree/main/skills/google-maps-platform ### Use Google Maps Platform agent skills #### Automatically trigger or manually invoke agent skills AI assistants are designed to use skills automatically whenever they detect that a skill's description matches your current request. However, skills can also be manually invoked. This is often done by typing `/` in the agent chat and searching for the skill name. For example, after installation, you can type the following in your agent chat to request a new feature in the application you're building: /google-maps-platform Add a 3D map featuring a marker for the user's location #### Use skills alongside the Code Assist MCP server Agent skills are designed to complement the [Code Assist MCP server](https://developers.google.com/maps/ai/code-assist.md.txt): - **Google Maps Platform agent skills**: Provide the quick instructions and recommended practices that tell an agent how to perform Google Maps Platform tasks through token-efficient progressive disclosure. They educate the agent on how to use the MCP server tools effectively. - **Code Assist MCP server**: Designed for AI-assisted development workflows, enabling AI assistants to interact with your Google Maps Platform projects, resources, and data programmatically. > [!TIP] > **Tip:** When you have both the Code Assist MCP server and Google Maps Platform agent skills installed, skills can instruct models how to use the MCP tools to accomplish complex tasks efficiently. > [!IMPORTANT] > **Important:** Agent skills are actively developed. To ensure your AI assistant always has the latest Google Maps Platform features, best practices, and security recommendations, be sure to regularly [update your skills](https://developers.google.com/maps/ai/agent-skills.md.txt#update-skills). ### Update Google Maps Platform agent skills You can get all available updates and clean up out-of-date skills by running the `update` command: ### Skills CLI ```bash npx skills update google-maps-platform ``` ### Gemini CLI ```bash gemini extensions update ``` ## Core components of a skill Each agent skill is a specialized package that can provide the AI assistant with the following components: - **Specialized instructions**: Detailed guidance on achieving specific tasks, such as implementing Place Autocomplete or generating a map with Advanced Markers. - **Best practices**: Built-in security and performance patterns to ensure your app follows Google Maps Platform recommendations from the start, including cost optimization. - **Automation scripts**: Executable code that allows your AI agent to perform local environment setup or configuration automatically. - **Real-world robustness**: Error and exception handling patterns distilled directly from verified open-source repository implementations. ## Benefits of using skills Using agent skills reduces the manual effort of searching documentation while improving the efficiency of your AI interactions. ### Reduced token costs Conventional AI integrations often load massive amounts of documentation upfront, which consumes significant tokens and increases session costs. Agent skills use **progressive disclosure** to minimize this overhead: - The agent initially only "scans" brief metadata to see if a skill is relevant. - Detailed instructions and resources are only loaded when the agent determines they're necessary for your specific task. ### AI-guided implementation Instead of manually searching through documentation, you can describe your intent in natural language. Skills guide your AI assistant to perform tasks such as the following: - *Add a map with a marker to my web app.* - *Implement Place Autocomplete for address entry.* - *Help me generate a static map image.* For a full list of the Google Maps Platform agent skills, see [Available skills](https://developers.google.com/maps/ai/agent-skills.md.txt#available-skills). ## Next steps - [Learn how to use agent skills with Gemini CLI.](https://geminicli.com/docs/cli/skills/) - [Learn how to use agent skills with Antigravity.](https://antigravity.google/docs/skills)