AI coding assistants promise to accelerate development, but often lead to frustration when they generate code based on outdated information or non-existent APIs. To solve this problem, we're releasing the Google Maps Platform Code Assist toolkit, an open-source Model Context Protocol (MCP) server designed to help your AI coding assistant generate accurate, optimal code grounded in the latest documentation, official code samples, and expert-authored best practices.
What is the Code Assist toolkit?
The Code Assist toolkit is an MCP server that provides your favorite AI coding assistant with up-to-date, official Google Maps Platform documentation, code samples, and best practices. By grounding your AI assistant in our official resources, it can generate more accurate, reliable, and useful code. You'll have more confidence and lower tech debt in the parts of your codebase that use Google Maps Platform, which is not what people usually say about the output of AI-generated code. The Code Assist toolkit gives your AI assistant access to the latest information from:
Google Maps Platform documentation
Google Maps Platform Terms of Service
Google Maps Platform Trust Center
Official Google Maps Platform GitHub repositories
Let's compare the responses of a coding assistant to a typical developer prompt, "Build me an address form that minimizes typos and provides visual confirmation of the entered address." Providing this prompt to several popular coding assistants powered by various coding models often yields implementations based on the Legacy version of the Places library for MapsJavaScript API because of the massive amount of training data from references around the internet based on long-lived documentation. With the Code Assist toolkit installed, the AI assistant gets context about the Places UI Kit component library that serves the same purpose with a familiar Google Maps user experience with minimal code and lower cost.
Without Code Assist toolkit
Example of an agent's response without Google Maps Platform Code Assist
With Code Assist toolkit
Example of an agent's response with Google Maps Platform Code Assist
How your AI assistant gets the latest information
The Code Assist toolkit acts as a bridge between your MCP client (like those in Gemini Code Assist, Cline, Cursor, and more) and a Google-hosted service that provides the latest contextual information.
With up-to-date context available on demand, your AI assistant's knowledge is no longer limited by its last training date. It can now access fresh, authoritative information maintained by Google. Since the Large Language Models (LLMs) powering AI assistants operate with a limited context capacity, the toolkit is designed to provide just the context needed to answer your prompt, ensuring efficient and relevant responses. The MCP server acts as an on-demand source of expertise for your AI coding assistant of choice, with more specialist understanding of how to apply Google Maps Platform developer resources than general-purpose grounding tools can provide.
Get started with Code Assist toolkit
The alpha version of the Google Maps Platform Code Assist toolkit is available now on GitHub and npm. If you’re using Gemini CLI, you can install Code Assist as an extension by running this command:
gemini extensions install https://github.com/googlemaps/platform-aiAfter installing Code Assist toolkit, when you prompt your AI assistant for help with a task involving Google Maps Platform, it will use tools from the Code Assist MCP server to retrieve documentation and code samples to ground the response otherwise known as retrieval-augmented generation (RAG).
If you’re using any other MCP client e.g. Cursor, Claude Code, etc. you can install the Code Assist toolkit using instructions at the documentation or the GitHub README. Once installed, your AI coding assistant will add the tool name and description from the Code Assist MCP server to its list of available tools.
With the Code Assist toolkit, you can code with more confidence knowing that your AI assistant's output is based on the latest guidance from Google Maps Platform.
