Screenpipe is a local-first app that continuously records your screen and audio into searchable history on your own device and opens that history to any AI. It is built for people who reconstruct their workdays and AI-assistant users who want grounded answers, keeping capture, storage, and search local with a REST API and MCP server instead of a cloud service.[1][2]
Quick verdict
People who lose things they saw or heard on screen and AI-assistant users who want answers grounded in actual work will get the most from Screenpipe. It records screen text, audio, apps, and URLs into encrypted on-device history and exposes it through search, a local API, scheduled tasks, and an MCP server. The price is a commercial source-available license for business use plus the ongoing curation that always-on capture demands, so open-source purists and teams needing SSO should look elsewhere. If you want your computer history searchable and usable by any AI, choose Screenpipe; if you only need simple automatic time tracking, choose ActivityWatch.[1][2][3]
Key features
Continuous screen and audio capture: Screenpipe records screen text through accessibility APIs with OCR fallback plus local audio transcription, along with app names and browser URLs, while it is running.[1]
Local-first encrypted history: recordings live in ~/.screenpipe on the device with zero-knowledge encryption, and per-app and URL exclusions plus on-device password and PII redaction run before anything is written to disk.[1]
Natural-language and API search: ask what you saw or heard in the app or query the local REST API on localhost:3030 with time, app, URL, speaker, and content-type filters.[1][2]
MCP memory for AI tools: an MCP server gives Claude, Cursor, ChatGPT, and other assistants searchable context from screen history, plus an activity-summary endpoint for compact agent readouts.[1][2]
Scheduled AI tasks: installable agents produce day recaps, standups, time breakdowns, meeting notes, and Obsidian syncs from recorded history.[1]
Dozens of app connections: settings wires Slack, Notion, Google Calendar, Obsidian, Toggl, HubSpot, Salesforce, Zoom, PostHog, Sentry, and more into capture and automations.[1]
Desktop apps on three platforms: official installers for macOS, Windows, and Linux, a terminal CLI via npx, and a self-maintained source build path.[1]
Source-available commercial license: the repository is inspectable and modifiable, free for personal, non-profit, educational, research, and 7-day evaluation use, with paid licenses required for commercial use and official builds under separate subscription terms.[1][2]
Who will love it
People who constantly lose things they saw or heard: if your day is reconstructing a meeting quote, an invoice step, or a page you closed, searchable screen and audio history answers from sources instead of memory.[1]
AI-assistant users who want grounded answers: piping recorded work context into Claude or ChatGPT through MCP turns vague chat into questions answered against what actually happened on screen.[1]
Individuals and small teams hunting automatable work: workflow discovery maps recurring cross-app runs with per-step source evidence, so the automation review starts from observed repetition.[1][2]
Who should avoid it
Businesses expecting an open-source license: the current license is a commercial source-available license, so any commercial use needs a paid license and official builds need a subscription; only legacy MIT-versioned code stays MIT.[1]
Teams needing SSO and shared administration today: no SSO documentation was found, licenses are per person with up to four inside one company, and five or more users need a Team or Enterprise plan.[1][2]
People unwilling to curate always-on capture: continuous recording means configuring app and URL exclusions, holding local storage, and choosing the AI egress mode deliberately, or the privacy posture is weaker than promised.[1]
Alternatives and compatibility
ActivityWatch covers simpler local activity history: it tracks apps, windows, and tabs without pixels or audio, so time-tracking purists can skip Screenpipe's heavier capture and AI surface.[1]
No vendor-endorsed catalogue replacement is claimed: Screenpipe's own README names Rewind, Limitless, Microsoft Recall, Granola, and Otter.ai as the products it positions against, but none of them exists in this catalogue, so no edge to them is proposed.[1]
MCP and scheduled tasks compose with existing tools: history flows into editors and assistants over MCP and into notes, calendars, and trackers through connections, so Screenpipe extends rather than replaces that stack.[1]
Questions worth asking
How much does Screenpipe cost per month?
Official builds are sold by subscription through the Screenpipe checkout with account entitlement, and trials may require payment details, so confirm the current price, renewal, and cancellation terms on the onboarding page before paying. Building from source is free for personal non-commercial, non-profit, educational, research, and 7-day evaluation use, while any commercial use needs a paid commercial license. AI provider access is billed separately under your own provider account. Prices checked October 12, 2026.[1][2]
Can you self-host Screenpipe?
There is nothing to host: Screenpipe runs as a local desktop app, CLI, or source build and keeps capture, processing, and storage on the device in ~/.screenpipe, with no account or server required for the core loop. Optional sync and enterprise archives write to storage you own rather than to Screenpipe servers. It runs on macOS, Windows, and Linux.[1][2]
Does Screenpipe work offline?
Yes in local-model mode: capture and search run on the machine and nothing leaves the device, with on-device encryption and per-app and URL exclusions plus PII redaction. Cloud transcription, hosted AI, provider keys, and connected apps are optional paths that send selected context off-device. Check the privacy data-flow page before enabling them.[1][2]
What is the best Screenpipe alternative for simple time tracking?
ActivityWatch is the simpler pick when you want automatic local-first time tracking without screen pixels or audio: it records active apps, windows, and browser tabs into an on-device database as free open-source software. Choose Screenpipe when you need full screen and audio recall plus AI search and agent context over that history. A catalogue relationship between the two is in review, not yet published.[1][2]
Sources and editorial review
Source research. Reviewed on October 12, 2026. Written by Sanjib Sah. AI assisted; reviewed by a human editor.
We have not run or tested this product.
- Screenpipe homepage
official · Observed October 12, 2026
Official site; local computer history open to any AI, YC-backed, individual and small-team positioning
- screenpipe/screenpipe on GitHub
official · Observed October 12, 2026
Canonical repo; YC S26, stars and forks, source-available positioning against Rewind, Recall, Granola, Otter.ai
- Screenpipe Commercial License
official · Observed October 12, 2026
Source-available terms; free personal and evaluation use, paid commercial use, MIT legacy versions
- Screenpipe security page
official · Observed October 12, 2026
Local-first storage and encryption, exclusions and redaction, AI egress modes, MDM and DPA
- Screenpipe documentation home
official · Observed October 12, 2026
Capture, search, scheduled tasks, MCP server, and app connections
- Screenpipe installation guide
official · Observed October 12, 2026
Desktop, CLI, and source-build paths with platform packages and license and payment terms
- Screenpipe API recipes
official · Observed October 12, 2026
Local REST API on localhost:3030 with search, summary, speaker, and retention endpoints
- ActivityWatch homepage
official · Observed October 12, 2026
Official site of the simpler local-first automatic time and activity tracker
Compared with these products
No explicit proprietary baseline is recorded.