screenpipe
Capture screen, audio, and app activity locally to provide AI agents with contextual memory. Works across macOS, Windows, and Linux with privacy by default.
screenpipe is a local-first tool that continuously captures your screen, audio, and app activity to create a searchable history for AI agents. It runs entirely on your device, keeping all data private while exposing a local API and MCP server for seamless integration with tools like Claude, Codex, and other agents.
Key Features
- Local Capture: Records screen frames, audio transcripts, and app activity directly on your device
- Privacy-First: All data stays on your machine with optional redaction of sensitive content
- Cross-Platform: Works on macOS, Windows, and Linux
- Agent Integration: Exposes history via local API and MCP server for any approved AI agent
- Searchable History: Find anything you've seen or heard with natural language queries
- Automated Workflows: Generate meeting notes, CRM updates, and follow-ups automatically
- Portable Data: Access your history through SQLite database, REST API, or inspectable source
- Enterprise Control: Manage devices, policies, and audit trails from a central workspace
Tech / Self-Hosting
- Open Source: Full source code available on GitHub
- Local-First Architecture: No cloud uploads by default (optional enterprise cloud sync)
- Performance Optimized: Minimal CPU, battery, and storage impact
- Extensible: Connect to existing tools via integrations (Notion, Slack, Linear, GitHub, etc.)
- Self-Hostable: Can be deployed across company devices with centralized management