screenpipe for Gemini CLI

Give Gemini CLI access to private, local-first workflow memory. The extension combines the screenpipe MCP server with an evidence-first skill for searching work history, reconstructing multi-app tasks, drafting cited SOPs, and finding automation candidates.

Requirements

• screenpipe desktop installed and running
• Node.js 18 or newer

The MCP server discovers the local screenpipe API key from the installed app. Package usage and error telemetry are disabled by this extension.

Install

gemini extensions install https://github.com/screenpipe/gemini-cli-extension

Screen and audio history stays in the user's screenpipe instance. Evidence retrieved for a task enters the configured Gemini model context. The bundled skill treats retrieval as the default and requires explicit user intent before mutations or externally visible actions.

The extension wrapper is MIT licensed. The screenpipe application and screenpipe-mcp package retain their respective upstream licenses.