screenpipe: what it is, what problem it solves & why it's gaining traction

What it solves

screenpipe creates a searchable, AI-powered memory of your computer activity. It solves the problem of forgetting what you've seen, heard, or done on your screen, providing a local-first, private alternative to cloud-based screen recording and memory tools.

How it works

• Event-driven capture: Instead of continuous recording, it captures screenshots and the OS accessibility tree (structured text) only when meaningful events occur (e.g., app switches, clicks, typing). It falls back to OCR when accessibility data is unavailable.
• Audio processing: It records system audio and microphone input, using local Whisper or cloud Deepgram for real-time speech-to-text transcription.
• Local storage: All data is stored in a local SQLite database with full-text search (FTS5), ensuring data remains on the device.
• AI Integration: It provides a REST API and an MCP (Model Context Protocol) server, allowing AI assistants like Claude or Cursor to query your screen history for context.
• Pipes: These are scheduled AI agents defined in markdown files that can automate tasks based on your activity (e.g., summarizing meetings or creating daily recaps).

Who it’s for

• Knowledge workers and researchers who need to recall specific information from their screen history.
• Developers who want to provide AI coding assistants with real-time context of their work.
• People with ADHD who seek a tool to track tabs, documents, and conversations.
• Remote workers who need automatic meeting transcriptions and notes.
• Enterprises looking to deploy AI agents across a team with deterministic, OS-level data permissions.

Highlights

• 100% Local: Data is stored locally by default, and supports local AI models via Ollama.
• MCP Server: Seamlessly integrates with AI assistants like Claude Desktop and Cursor.
• Deterministic Permissions: Per-pipe YAML-based access control prevents AI agents from accessing unauthorized data.
• Multi-platform: Supports macOS, Windows, and Linux.
• Low Overhead: Uses 5-10% CPU and event-driven capture to minimize storage and disk usage.