Overview:
Screenpipe is an open-source application that continuously captures your screen and audio to create a searchable, AI-powered memory of your computer activity. All data is stored locally on your device. It targets knowledge workers, developers, researchers, remote workers, and teams who need to recall and search through past on-screen information, audio conversations, and work context without relying on cloud-based services. It runs on macOS, Windows, and Linux.
Core Features:
Event-driven screen capture: Captures screenshots and accessibility tree data only when meaningful system events occur (app switches, clicks, typing pauses, scrolling), falling back to OCR when structured text is unavailable.
Local audio transcription: Records system and microphone audio, performing real-time speech-to-text via a locally-run OpenAI Whisper model with speaker identification and diarization.
AI-powered search: Enables natural language and semantic search across all captured text and transcriptions, with filters by application, window title, URL, and date range.
Plugin system (Pipes): Allows users and developers to create scheduled AI agents as markdown files that query screen data, call APIs, and execute actions, with built-in pipes for Obsidian sync, reminders, and idea tracking.
Developer API and MCP server: Provides a full REST API on localhost and a Model Context Protocol server, enabling other AI assistants like Claude, Cursor, and Cline to directly query screen history.
Team deployment features: Supports central configuration management, shared AI pipes, and per-pipe YAML-based data permissions enforced at the OS level for deployed organizations.
Use Cases:
Knowledge workers recalling past work: Search through past screen content, meetings, or documents using natural language to quickly find specific information or conversations.
Developers providing context to AI coding assistants: Give tools like Cursor, Claude Code, or Continue access to recent screen activity and meeting transcriptions to improve their awareness of current tasks.
Researchers searching large volumes of on-screen information: Browse and query a chronological timeline of screen captures and audio to track research sessions, locate specific references, or review past interactions.
Remote workers getting automatic meeting notes: Capture and transcribe audio from any meeting application (Zoom, Google Meet, Teams) to generate searchable notes and summaries without manual recording.
Why It Matters:
Screenpipe is a self-contained, open-source alternative to cloud-based screen memory tools. It operates entirely locally by default, with optional encrypted sync between devices. Its event-driven capture reduces CPU and storage load compared to continuous recording, and its team framework provides deterministic, non-prompt-based data permissions for AI agents. Unlike comparable closed-source solutions, it supports multiple platforms, local AI model choice, and a plugin architecture for custom automation.
