Screenpipe
Screenpipe is a source-available application that continuously captures your screen and audio locally, creating a searchable, AI-powered memory of your computer activity. It solves the problem of forgetting what you've seen or heard by providing context to AI agents and enabling natural language search. It is for knowledge workers, developers, researchers, and anyone wanting a private, local-first alternative to cloud-based AI memory tools.
✨ Key features
- Event-driven screen capture with accessibility tree and OCR fallback
- Real-time audio transcription with speaker diarization (local Whisper or Deepgram)
- AI-powered natural language search across screen text and audio
- Timeline view to scroll through your screen history like a DVR
- Plugin system (Pipes) for scheduled AI agents defined as markdown files
- MCP server for integration with AI assistants like Claude and Cursor
🎯 Use cases
- Recall what you saw or heard on your computer with natural language queries
- Give AI coding assistants context about your recent work
- Automatically generate meeting summaries and action items
- Track time and productivity across apps and projects
- Create custom automations that react to your screen activity
📦 Installation
🧰 Requirements: macOS (Apple Silicon or Intel), Windows 10/11, or Linux; 8 GB RAM recommended; ~5-10 GB disk space per month; no account required for core app.
Download the desktop app from screenpipe.com for all features and auto-updates.
Or run the CLI:
npx screenpipe record
Then set up the MCP server:
npx screenpipe setupclaude mcp add screenpipe -- npx -y screenpipe-mcp@latest🚀 Usage
After installation, ask your AI assistant (e.g., Claude) questions like:
what did i see in the last 5 mins?
Or run a pipe to generate a daily recap:
npx screenpipe setup
Then ask Claude to "summarize today conversations" or "create a pipe that updates linear every time i work on task X".
⚠️ Good to know
The signed desktop app requires a subscription (Standard $25/mo, Pro $50/seat/mo, Enterprise $150/seat/mo); source is available for personal, non-commercial use only.
❓ FAQ
Is my data stored locally?
Yes, all data is stored locally on your device in a SQLite database by default. Nothing is sent to external servers unless you opt into cloud features.
What platforms are supported?
macOS (Apple Silicon and Intel), Windows 10/11, and Linux (build from source) are supported.
How much CPU and storage does it use?
CPU usage is typically 5-10% on modern hardware, RAM usage is 0.5-3 GB, and storage is about 20 GB per month.
Can I use it with my own AI models?
Yes, you can use local AI models like Ollama or any OpenAI-compatible server, keeping data fully local.
📊 Repository
🤖 Overview, features, install steps and FAQ were generated from the project's README on Sep 4, 2026. Always check the original source before running commands.