screenpipe

Screenpipe runs in the background on your machine, recording your screen and microphone around the clock. Everything stays local. No cloud upload, no third-party storage. When you need to know what was said in a meeting last Tuesday, or who asked about pricing last month, you ask an AI and it answers from your own captured history.

It's a direct alternative to tools like Rewind.ai but built on a fully open, self-hostable foundation. The recording and transcription happen on-device, and the AI queries run against your local memory store.

What it captures and does:

• Continuous screen recording across all apps, stored and indexed locally
• Microphone transcription for meetings, calls, and any spoken audio
• AI querying against your full history using models like Claude, Qwen, or local options via Ollama
• Pipes – lightweight automations that trigger on events like a meeting ending, then write notes, update a CRM, or post to Slack
• 48+ integrations including Linear, GitHub, HubSpot, Notion, Google Docs, and Zoom

The Pipes system is where it gets practical. You write a short instruction file describing what should happen when a trigger fires. A meeting ends, the pipe transcribes it, summarizes it into bullet points and action items, and pushes the result to wherever you want. No manual copy-paste, no switching apps. Tools like Meetily or Hyprnote handle meeting notes in isolation; Screenpipe connects that output directly into the rest of your workflow.

It runs on macOS, Windows, and Linux. The AI layer is model-agnostic, so you can swap between hosted models and fully local ones depending on your privacy requirements. For anyone who's tried dedicated meeting transcription tools and wanted something that goes further than just notes, Screenpipe treats your entire desktop activity as queryable memory.