Stitch Extension for Gemini CLI MCP
A Gemini CLI extension for Google Stitch: the /stitch command lists projects and screens, downloads images and HTML and generates screens from text
- 479 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- The extension calls a remote Google service and sends it product descriptions and generation requests
- The API key or Google credentials open access to your Stitch projects and screens
Install
- Works in
Manual install
gemini extensions install https://github.com/gemini-cli-extensions/stitch --auto-updateGemini CLI 0.19.0 or newer is required. Then set up sign-in with an API key or ADC.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The extension calls a remote Google service and sends it product descriptions and generation requests
- The API key or Google credentials open access to your Stitch projects and screens
This is third-party code. Review the repository files before installing.
What it does
The extension connects Gemini CLI to the Google Stitch MCP server, an AI service for interface design and code generation. After installation a /stitch command appears in the terminal, and you can write to it in plain phrases. The agent lists your Stitch projects, opens the details of one by its number and lists all its screens. From a screen you can download an image or an HTML file. New screens are made from a text description with Gemini 3 Flash by default or Gemini 3 Pro if you name it in the request. A prompt enhancement mode exists too: the extension rewrites a sketch like "landing page for a podcast about design and AI" into a fuller description. The server is remote and free of charge, answering at https://stitch.googleapis.com/mcp. You can sign in with an API key from the Stitch settings or through Application Default Credentials in your own Google Cloud project.
Who it is for. For designers and frontend developers who work in Google Stitch and want to manage projects and screens from the terminal through Gemini CLI.
Good fit when
- You want to see all Stitch projects and the screens of one of them fast, without the web interface
- A design is ready in Stitch and its HTML or image has to land in a repository for implementation
- You want to sketch a mobile app from a text description and choose Gemini 3 Pro for a complex screen
- A short landing page description should become a detailed prompt: the prompt enhancement command does that
- You need API key access for personal use without setting up a Google Cloud project
Not a fit when
- You work in Claude Code or Cursor: the extension is built for Gemini CLI 0.19.0 and newer, and Stitch has separate skills for other agents
- You need pixel-precise manual editing of a design: the extension reads, downloads and generates screens and is no editor
- You need screens with no network access: the server is remote
- Company policy forbids sending product descriptions to Google services: requests go to Stitch
- You need ready React components: the extension returns screens, HTML and images
Example request
/stitch Show my projects, take the screens of the ski app project and download the HTML of the first screenMCP
- Transport
- http
- Authentication
- API key
Environment variables
API_KEYrequired, secret- The API key from Stitch settings, substituted in the extension for YOUR_API_KEY
Limitations
The extension works only in Gemini CLI version 0.19.0 or newer. Sign-in is set up by hand in one of two ways. With an API key from the Stitch settings, you put the key into ~/.gemini/extensions/Stitch/gemini-extension.json with a sed command. With Application Default Credentials you need your own Google Cloud project: sign in through gcloud, enable the stitch.googleapis.com service, grant the roles/serviceusage.serviceUsageConsumer role and substitute the project ID.
The feature set is still small: project list, details, screens, download and generation. The author promises more later. Use is governed by Google's terms, your Stitch settings and the Stitch privacy notice. The author does not say whether Stitch is reachable from Russia.
How to disable. Remove the extension with gemini extensions uninstall Stitch. The API key you put into gemini-extension.json goes away with the extension folder. You can revoke the key itself in the Stitch settings, API Keys section.
FAQ
How do I install the Stitch extension in Gemini CLI?
Run gemini extensions install https://github.com/gemini-cli-extensions/stitch --auto-update. Then set up sign-in: put the API key from the Stitch settings into gemini-extension.json or go through Application Default Credentials. After launching gemini, check the commands with /mcp list.
Is Stitch MCP paid?
No, Stitch MCP is free of charge. Terms of use are set by the Google Terms and Use Policy, the Google APIs terms and your Stitch settings.
Which sign-in should I pick, an API key or ADC?
An API key is simpler: create it in Stitch under Stitch Settings, then API Keys. The author advises it for personal use and a quick start. ADC fits when you work in your own Google Cloud project, but takes several manual steps and the serviceUsageConsumer role.
Which models generate the screens?
Gemini 3 Flash by default. To use Gemini 3 Pro, name it in the request, for example "using Gemini 3 Pro".
How does the extension differ from Stitch skills for other agents?
The extension connects Gemini CLI to the Stitch MCP server and works through the /stitch command. Stitch skills from a separate repository install as plugins and help agents generate screens and move code into design.
Discussion
No questions or reviews yet.
Related
MCP serversThree skill plugins for Google Stitch: screen generation, code to design and React builds from mockups
Story UI
story-ui
Storybook story generation with AI for any framework and design system, with an MCP server. An Anthropic
Figma's official MCP server: code from a selected frame, design system variables and components, and writing designs to the canvas
An MCP server that gives the agent simplified Figma layout data from a link so it can build the screen more accurately than from a screenshot
An MCP server and plugin to build UI with the agent: curated design systems, UI prompts and layout variations
An MCP server and skill for interfaces: 31 design systems for shadcn/ui, UI and UX principles and accessibility review for agent-written code
README in short
The document lists what the extension can do: project list, details, screens, asset download and screen generation from text. Then come prerequisites, a single install command and two sign-in methods with step-by-step instructions for an API key and for Application Default Credentials. The usage section shows request examples with the /stitch prefix, including asset download and prompt enhancement. At the end it notes that the service is free and gives resource links and Google's legal terms.