Input

You can use the input-mcp server to collect contextual user input through a single dedicated tool called collect_input. It opens an Electron-based prompt for text, image, or pixel art input and returns the result to your MCP client. This makes it easy to gather user-created content in a consistent, interactive way within your workflow.

How to use

To collect input from users, call the collect_input tool from your MCP client. You can request different input kinds depending on your needs:

• Text input: capture simple text or JSON-like data. - Image input: allow users to draw or annotate on a canvas. - Pixel art: provide a grid-based editor with a color palette for compact artwork.

When you invoke collect_input, the tool launches a window for the user to provide their input. After the user completes their input, you receive the resulting data back in your MCP client. For image inputs, the result is typically a file path to the saved image.

Practical usage ideas include gathering feedback with text, collecting quick sketches for UI concepts via image input, or creating simple pixel art assets for games or icons. Design your client flows to handle all three input kinds by passing the appropriate spec to collect_input.

How to install

Prerequisites you need before installing:
- Node.js and npm (or a compatible runtime)
- A working shell to run commands
- Optional: Bun for faster tooling in local development

npm install
npm run build
npx @swairshah/input-mcp

Additional setup and testing

If you want to smoke-test locally, you can run the server and use the built-in inspector to verify behavior. Build the inspector-enabled server and start it with the inspector wrapper.

npm install
npm run build
npx @modelcontextprotocol/inspector node dist/src/server.js

Using with a client like Claude

Configure your client to start the MCP server by executing the stdio command shown above. You can also run the MCP package directly via npx to start a local server instance.

Examples of input kinds

• Text input example: invoke collect_input with kind set to text to capture free-form text or JSON payloads.

• Pixel art example: invoke collect_input with kind set to pixelart and specify grid dimensions to open a pixel art editor.

• Image example: invoke collect_input with kind set to image, optionally providing an initial image to edit.

Available tools

collect_input

Collects contextual user input by launching an Electron window. Supports text input, image input with optional initial image, and pixel art input with a grid-based editor.