Screenpipe

After you install Screenpipe on macOS, Windows, or Linux, it captures screen frames with OCR, system and microphone audio transcripts, and accessibility context, and writes them to a local database on your machine. You control retention and can exclude specific apps, windows, hours, or URLs so sensitive material never lands in the store. Approved AI tools reach that history through a local REST API (for example a search endpoint on localhost) and an MCP server, which lets Claude, Codex, or a custom agent ask grounded questions like what happened in the last five meetings with a given account. You can also install scheduled workflows (pipes) that trigger on events, such as writing a summary and action items when a meeting ends and pushing them to your notes.

• •Local-First Capture: Records screen, audio, and app context on-device with configurable retention, so your raw work history never has to leave your computer.
• •Local API & MCP Server: A REST API and MCP server expose your history to any approved AI tool, keeping your memory portable across models instead of tied to one vendor.
• •Scheduled Agent Workflows: Install pipes that trigger on events (meeting ended, and more) to draft summaries, update trackers, or push follow-ups automatically in the background.
• •Privacy Controls & Redaction: Exclude apps, windows, URLs, and hours, and redact sensitive fields on-device before anything is stored.
• •Cross-Device Sync & Team Workspaces: Paid tiers add synced history across devices, managed team seats, and higher-throughput models for questions about your work.
• •Open and Inspectable: The source is available on GitHub, so you can audit exactly what is captured and how it is used.

Screenpipe fits knowledge workers who live in their computer all day and want their AI tools to actually remember it. Founders and salespeople use it to reconstruct deal context and draft follow-ups from real calls. Consultants and researchers use it to turn scattered work across apps into searchable, reusable memory. Developers and AI builders are the natural core audience: the local API, MCP server, and open source make it straightforward to wire your own agents into a private, on-device context store. Privacy-conscious teams who cannot send raw activity to a third-party cloud can deploy it with local storage and device controls, and enterprises get SSO, MDM, retention, and audit controls on managed rollouts. If you want durable AI memory without surrendering your data, Screenpipe is built for you.