Why it exists

You saw the exact setting that fixed a bug last week, heard the decision in Thursday's meeting, copied the same invoice data into two systems all month — and remember none of it by Friday. Browser history and file search only cover fragments. Screenpipe's pitch is a continuous, local record of your actual computer work that turns "I vaguely remember doing something like this" into a query — and then lets AI act on it: draft the follow-up, write the how-to, or surface that you're doing the same manual copy-paste every single day.

Main features

• Continuous local capture — screen, audio, and meetings recorded on-device; raw history stays local by default.
• Ask AI about your work — "pick up where I left off", "reconstruct my day", "map how I processed invoices this week" — answers grounded in your real history instead of the model's imagination.
• Workflow discovery — finds recurring patterns across apps (e.g., invoice email → accounting app → tracker) and flags steps that could be automated, with approval kept with you.
• Works with any AI — share selected context with Claude, Codex, and other tools; history is exposed through MCP so agents can search your work.
• Scheduled workflows — automations grounded in what was actually captured, installed and run on a schedule (paid tiers).
• Team layer — managed seats and cross-device sync of your own context on Business; note shared teammate context is not included.

Who it's for

• Power users who want a "photographic memory" of their computer — what they saw, said, and clicked.
• People automating their own busywork: the discovery view is built to surface the repetitive processes you'd never think to script.
• Developers wiring personal context into Claude/Codex/agent setups via MCP.
• Small teams trying to document tribal knowledge — how work actually happens — from real captured history.

How it compares

• Rewind/Limitless (formerly Rewind AI) — the closest rival in Mac-focused screen memory; Screenpipe differentiates with cross-platform capture, an open(ish) engine, MCP integration, and a workflow-automation angle rather than pure recall.
• Manual documentation (SOPs, Loom) — accurate but instantly stale; Screenpipe derives the steps from what you actually did, then you clean it up.
• Plain OS activity history — browser and file recents are thin, siloed fragments; Screenpipe unifies screen + audio + app context into one searchable store.
• Hister / personal search engines — index what you choose to save; Screenpipe captures what you did, no filing discipline required.

Notes

• The privacy story depends on the local-first default: capture stays on-device on the Free tier; sync across devices is a paid feature. On a shared or employer machine, decide deliberately whether this should run at all.
• Licensing is unusual: the GitHub repo is public (21.9k stars) but under a custom commercial license — free for individuals on their own devices; business/production use requires a paid plan (5+ users at one company need Team/Enterprise).
• Pricing: Free (one device, recent history, MCP sharing), Basic $21/mo (full searchable history, unlimited scheduled workflows), Business $42/seat/mo (sync, managed seats).
• Backed by Y Combinator (S26); used at Google, Microsoft, NVIDIA, MIT, Shopify and others per the site's own banner.
• Constant capture is power-hungry and disk-hungry — check the local retention settings before trusting it on a laptop.

Useful links

• Official site: https://screenpipe.com/
• Download: https://screenpipe.com/download
• Pricing: https://screenpipe.com/pricing
• GitHub: https://github.com/screenpipe/screenpipe
• Security page: https://screenpipe.com/security