Screenpipe records your screen and audio locally so AI agents can search what you’ve seen and heard

Launch HN: Screenpipe (YC S26) – Power your agents by your 24/7 screen recording

Screenpipe (YC S26) is a local screen and audio recorder that gives AI agents searchable context about what happened on your computer. Instead of recording full video, it listens for app switches, clicks, typing pauses, and scroll events, then pairs screenshots with the OS accessibility tree; OCR is only used when structured data is missing. Audio is transcribed locally via Parakeet/Whisper. Everything is stored in a local SQLite database and served through an authenticated API with MCP and skills support, so agents like Claude or ChatGPT can retrieve past tasks, generate daily summaries, maintain a personal wiki, or spot automation opportunities. A local PII redaction model runs on Apple MLX or Windows DirectML, and users can set app/window/URL filters plus recording schedules. The code is source-available under a new commercial license—free for personal non-commercial use, paid for commercial. The post does not disclose pricing or latency figures.

Why it matters: Screenpipe gives agents a local memory layer with a smarter capture approach than dumb screen recording. But this is a Launch HN—product just debuted, no user scale data, no real-world deployment stories. Score sits right at the featured threshold.