Screenpipe
ActiveDescription
Screenpipe records your screen and audio locally around the clock, turning everything you do on your computer into a searchable memory that gives coding agents real context through MCP.
Key Features
- Event-driven capture - screenshots only when something meaningful happens, pairing each frame with the OS accessibility tree at 5-10% CPU
- Local audio transcription - Whisper runs on-device to transcribe system audio and microphone input with speaker identification
- Natural language search - query screen text, OCR output, and transcripts with filters for app, window title, URL, and date range
- MCP server - zero-config integration lets Claude Desktop, Cursor, and other assistants query your screen history directly
- Pipes plugin system - scheduled AI agents defined as markdown files produce meeting summaries, day recaps, and time breakdowns
- 100% local storage - everything lives in a local SQLite database with optional encryption at rest and full offline support
Use Cases
💡 Give coding agents real context about what you are working on so they can pick up tasks without repeated questions.
💡 Automatically summarize meetings after they end and patch the notes back onto the meeting record.
💡 Recall any screenshot, conversation, or document you saw days ago using natural language search.
💡 Run scheduled pipes that generate standup updates and time breakdowns without manual bookkeeping.
Strengths & Limitations
✅ Strengths
- • Actively maintained, recent updates
- • High community interest (21.5k stars)
- • Established track record (2 years in production)
- • Responsive to issues, low backlog
⚠️ Limitations
- • No clear open-source license
Categories
Quick Start
# Start the local recording service
npx screenpipe record
# Set up skills and MCP for every detected agent
npx screenpipe setup
# Or wire screenpipe into Claude Code manually
claude mcp add screenpipe -- npx -y screenpipe-mcp@latest