
Screenpipe is a local-first AI memory app for desktop work. It captures screen and audio activity and turns it into searchable memory, meeting notes, SOPs, reports, and automations for teams that want AI agents grounded in real work context.
Most AI memory tools ask users to trust a hosted recorder. Screenpipe records locally, stores memory on disk, and lets users drop sensitive apps, windows, and URLs before they reach storage. The company also publishes the recorder code and exposes open APIs, so teams can inspect the capture layer and query memory.
Screenpipe treats work memory as infrastructure, not just search. Pipes can trigger from events like a meeting ending, summarize what happened, and push structured output into the tools a team already uses.
The workflow is simple: install the desktop app, let it capture screen and audio, then search or ask AI about what happened. It can answer questions about meetings, pricing discussions, progress, account context, or past decisions using captured context.
For automation, pipes create meeting notes, CRM updates, follow-ups, time tracking, SOPs, deal spotting, and contact dossiers. Teams can connect more than 50 tools, including Notion, Slack, Linear, GitHub, Zoom, HubSpot, Salesforce, OpenAI, Claude, Ollama, and Gemini. Enterprise deployment adds managed seats, SSO or SAML, MDM controls, on-prem storage, retention, workflow reports, and custom agents.
No public average customer rating is presented. The company lists developer and community signals, including 19k GitHub stars and 3k Discord members.
The main tradeoff is resource use. Requirements list 600 MB to 4 GB RAM depending on settings, 10-30% CPU on Apple M3, similar behavior on modern Intel or AMD machines, and about 30 GB of storage per month with configurable retention.
The core is source-available and free to build or self-host for personal, non-commercial use. Paid plans cover the signed app, updates, support, sync, integrations, and team features.
Data stays on device by default, is encrypted at rest, and can be filtered by app, window, URL, plus on-device PII redaction.
It can use noticeable resources: requirements list 600 MB to 4 GB RAM, 10-30% CPU on Apple M3, and configurable storage.
The core is source-available and free to build or self-host for personal, non-commercial use. Monthly paid plans include a 7-day trial.
It fits people who need searchable work memory, meeting notes, and automations from local screen and audio capture.
Ask specific questions about this tool.
+2 more