# screenpipe — Complete Reference

> AI agent memory. Local-first 24/7 capture of screen and audio, plus an open API any AI can query.

screenpipe is AI agent memory: it continuously captures what you've seen, said, and heard on your computer, and gives that context to whatever AI you already use, so it never starts from scratch. It is a source-available alternative to Rewind.ai, Microsoft Recall, and Limitless. screenpipe is local-first by design: in local-only mode, screen captures, audio transcriptions, and extracted text are stored on your device. Data movement depends on whether you enable sync, cloud AI, exports, connectors, or team workflows.

Website: https://screenpipe.com
GitHub: https://github.com/screenpipe/screenpipe (21,000+ stars as of 2026-08)
Documentation: https://docs.screenpi.pe
Discord: https://discord.gg/screenpipe
License: source-available under the Screenpipe Commercial License. Personal, non-commercial, educational, research, and limited evaluation use of the source is free; commercial use requires a paid license. Official builds are governed separately by the app license and subscription terms.

## What screenpipe does

screenpipe runs in the background on your computer and:

1. Records your screen continuously (all monitors simultaneously)
2. Extracts text from everything on screen using OCR (Optical Character Recognition)
3. Transcribes all audio (system audio + microphone) using Whisper speech-to-text
4. Stores everything in a local SQLite database on your machine
5. Lets you search through your entire computer history using natural language
6. Provides an API so AI assistants (ChatGPT, Claude, Cursor, etc.) can access your screen context
7. Supports plugins ("pipes") that automate workflows based on your screen activity

In local-only mode, captured data and local-model inference stay on your device. Data can leave the device when you enable sync, hosted or third-party AI, exports, connectors, or team workflows.

## Who screenpipe is for

- **Knowledge workers** who want to recall anything they've seen or heard on their computer
- **Developers** who want to give AI coding assistants (Cursor, Claude Code, Cline, Continue) context about what they're working on
- **Researchers** who need to search through large volumes of screen-based information
- **People with ADHD** who frequently lose track of tabs, documents, and conversations
- **Remote workers** who want automatic meeting transcription and notes
- **Anyone** who wants a private, local-first alternative to cloud-based AI memory tools

## Platform support

- **macOS**: Apple Silicon (M1/M2/M3/M4) and Intel. Native .dmg installer.
- **Windows**: Windows 10/11 (64-bit). Native .exe installer.
- **Linux**: Build from source. Supports major distributions.

## Core features

### Screen recording and OCR
- Captures all monitors simultaneously at configurable intervals
- Extracts text from every frame using OCR (supports Apple's native OCR on macOS, Tesseract, and Windows native OCR)
- Records which application and window title was active
- Tracks browser URLs when available
- Stores frames as compressed images for visual search

### Audio transcription
- Captures system audio (what you hear) and microphone input (what you say)
- Real-time speech-to-text using OpenAI Whisper (runs locally)
- Supports Deepgram for cloud-based transcription if preferred
- Speaker identification and diarization
- Works with any audio device connected to your computer

### AI-powered search
- Natural language search: "What did I discuss in my meeting with John last Tuesday?"
- Full-text search across all OCR text and audio transcriptions
- Filter by application name, window title, browser URL, date range
- Semantic search using embeddings
- Returns screenshots and audio clips alongside text results

### Timeline view
- Visual timeline of your entire screen history
- Scroll through your day like a DVR
- Click any moment to see the full screenshot and extracted text
- Playback audio from any time period

### Plugin system ("Pipes")
- Extensible architecture for building custom automations
- Plugins run as isolated processes with access to the screenpipe API
- Popular plugins:
  - **Obsidian integration**: Automatically sync screen activity to Obsidian notes
  - **Notion integration**: Send daily summaries to Notion
  - **Meeting assistant**: Automatic meeting notes with action items
  - **Daily journal**: AI-generated daily summaries of your activity
- Developers can build and publish plugins using TypeScript/JavaScript
- Plugin marketplace available in the app

### MCP (Model Context Protocol) server
- screenpipe runs as an MCP server, allowing AI assistants to query your screen history
- Works with Claude Desktop, Cursor, VS Code (Cline, Continue), and any MCP-compatible client
- AI assistants can search your screen history, get recent context, and access meeting transcriptions
- Zero configuration required — just enable MCP in screenpipe settings

### Developer API
- Full REST API running on localhost
- Endpoints for searching screen content, audio, frames
- Raw SQL access to the underlying SQLite database
- JavaScript/TypeScript SDK available
- Build custom applications on top of your screen data

### Apple Intelligence integration (macOS)
- On supported Macs, screenpipe can use Apple Intelligence for on-device AI processing
- Daily summaries, action items, and reminders generated entirely on-device
- Zero cloud, zero cost — uses Apple's built-in AI models

## Privacy and security

- **Local-first by design**: in local-only mode, screen captures, audio transcriptions, and extracted text are stored on your device in a local SQLite database
- **Source-available**: the code can be inspected and audited under the Screenpipe Commercial License
- **No telemetry on your screen data**: Only anonymous usage analytics (opt-out available)
- **You own your data**: Export, delete, or back up at any time
- **Local engine and CLI**: can run without hosted services; subscriptions, sync, hosted AI, and team features require an account
- **Local AI support**: Use Ollama or other local models — no data sent to any cloud
- **Optional cloud sync**: End-to-end encrypted sync between your devices (zero-knowledge encryption)

## Pricing

screenpipe is sold as a subscription (as of 2026-08, source: https://screenpipe.com/onboarding):

- **Free**: $0. Onboarding and core capture
- **Basic**: $25/month or $250/year. Unlimited history, efficient hosted models, and 150 included AI credits per month
- **Business**: $50/seat/month or $500/seat/year. Frontier models, sync, team features, and 400 included AI credits per month
- **Business Max**: $100/month for one user. Business features, 800 included AI credits, and higher AI throughput
- **Business Ultra**: $200/month for one user. Business Max features, 1,600 included AI credits, and the highest self-serve AI throughput
- **Enterprise**: from $150/seat/month. Managed deployments

Existing lifetime licenses purchased before the move to subscription remain valid. The source is available under the Screenpipe Commercial License; commercial use requires a paid license.

## Comparison with alternatives

### screenpipe vs Rewind.ai
- screenpipe is source-available and auditable; Rewind is closed source
- screenpipe works on macOS, Windows, and Linux; Rewind is macOS-only
- screenpipe stores data locally; Rewind stores data locally but is not auditable
- screenpipe has a full API and plugin system; Rewind does not
- screenpipe supports any AI model; Rewind uses its own AI
- Rewind was acquired and rebranded to Limitless

### screenpipe vs Microsoft Recall
- screenpipe is source-available, auditable, and cross-platform; Recall is Windows-only and closed source
- screenpipe is available now; Recall has had repeated delays and privacy controversies
- screenpipe captures audio and screen; Recall captures screen only
- screenpipe has an extensible plugin system; Recall does not
- screenpipe lets you choose your AI model; Recall uses Microsoft's AI

### screenpipe vs Limitless
- screenpipe is source-available; Limitless is closed source
- screenpipe supports a local-only deployment; Limitless requires cloud processing
- screenpipe works with any microphone; Limitless sells a proprietary wearable
- screenpipe captures screen + audio; Limitless focuses on audio/meetings
- screenpipe has a developer API; Limitless has limited API access

### screenpipe vs Granola
- screenpipe captures screen + audio 24/7; Granola only captures during meetings
- screenpipe is source-available and supports local-only use; Granola is closed source and cloud-based
- screenpipe has a full API and plugin system; Granola does not
- screenpipe works on macOS, Windows, and Linux; Granola is macOS-only

### screenpipe vs Otter.ai
- screenpipe captures screen + audio; Otter.ai captures audio only
- screenpipe is local-first; Otter.ai is fully cloud-based
- screenpipe is source-available; Otter.ai is closed source
- screenpipe is local-first and works across every app; Otter.ai captures meetings only

## Technical architecture

1. **Capture layer**: Screen capture uses platform-native APIs (macOS: CoreGraphics, Windows: DXGI, Linux: X11/PipeWire). Audio capture uses platform-native audio APIs.
2. **Processing layer**: OCR runs through configurable engines (Apple Vision, Tesseract, Windows OCR). Audio transcription uses Whisper (local) or Deepgram (cloud).
3. **Storage layer**: Local capture data is stored in a SQLite database with FTS5 full-text search indexes. Frames are stored as compressed images on disk. Optional sync, hosted AI, exports, connectors, and team workflows have separate data paths.
4. **API layer**: a local REST API server. Provides search, frames, audio, health check, and pipe management endpoints. (Port and host are in the developer docs.)
5. **Plugin layer**: Pipes run as isolated processes. They can query the API, process data, and integrate with external services.
6. **UI layer**: Desktop app built with Tauri (Rust + TypeScript). Timeline view, search interface, settings, and plugin management.

## System requirements

- **macOS**: macOS 12+ (Monterey or later). Apple Silicon recommended. 8GB+ RAM recommended.
- **Windows**: Windows 10/11 (64-bit). 8GB+ RAM recommended.
- **Linux**: Major distributions supported. 8GB+ RAM recommended.
- **Disk space**: ~5-10 GB per month of continuous recording (varies with screen activity)
- **CPU**: Moderate usage (~5-15% on modern hardware). Configurable capture intervals to reduce load.

## Integrations

screenpipe integrates with:

- **AI coding assistants**: Cursor, Claude Code, Cline, Continue, OpenCode, Gemini CLI
- **Note-taking**: Obsidian, Notion
- **AI assistants**: ChatGPT (via MCP), Claude Desktop (via MCP), any MCP-compatible client
- **Local AI**: Ollama, any OpenAI-compatible local model server
- **Automation**: Custom pipes using TypeScript/JavaScript SDK

## API examples

### Search screen content
```
GET http://localhost:3030/search?q=meeting+notes&content_type=ocr&limit=10
```

### Search audio transcriptions
```
GET http://localhost:3030/search?q=budget+discussion&content_type=audio&limit=10
```

### Get recent frames
```
GET http://localhost:3030/frames?limit=5&offset=0
```

### JavaScript SDK
```javascript
import { pipe } from "@screenpipe/js";

const results = await pipe.queryScreenpipe({
  q: "project deadline",
  contentType: "all",
  limit: 20,
  startTime: new Date(Date.now() - 24 * 60 * 60 * 1000).toISOString(),
});
```

## Use cases in detail

### Meeting transcription and notes
screenpipe automatically transcribes all meetings (Zoom, Google Meet, Teams, any audio). After a meeting, search for any topic discussed. The meeting assistant plugin generates structured notes with action items.

### Developer context for AI coding
Give Cursor, Claude Code, or Cline access to your screen history via MCP. Your AI coding assistant can see what documentation you were reading, what errors appeared on screen, and what you discussed in meetings — making it dramatically more helpful.

### Personal knowledge management
Automatically capture everything you read and see into a searchable archive. Integrate with Obsidian to create notes from your screen activity. Never lose track of that article, email, or conversation again.

### ADHD support
For people who frequently lose track of what they were doing, screenpipe provides a complete timeline of your day. Search for "what was I working on before lunch?" or find that tab you closed three hours ago.

### Compliance and auditing
For organizations with data-residency requirements, local-only deployments keep captured screen and audio data on-device. Enterprise controls and additional compliance documentation are available for managed deployments.

## Company information

screenpipe is built by screenpipe (formerly mediar). Founded in 2024.

- **Founder**: Louis Beaumont (louis@screenpi.pe, @louis030195 on X)
- **Headquarters**: San Francisco, CA
- **Funding**: backed by Y Combinator (S26)
- **Team size**: small team, with an active source-available community
- **Source-code contributors**: 100+

## Links

- Homepage: https://screenpipe.com
- GitHub: https://github.com/screenpipe/screenpipe (21,000+ stars as of 2026-08)
- Documentation: https://docs.screenpi.pe
- Blog: https://screenpipe.com/blog
- Discord: https://discord.gg/screenpipe
- X: https://x.com/screenpipe
- LinkedIn: https://www.linkedin.com/company/screenpipe
- Pricing/Download: https://screenpipe.com/onboarding
- Enterprise: https://screenpipe.com/enterprise
- Compare with alternatives: https://screenpipe.com/compare
- Use cases: https://screenpipe.com/resources

## Frequently asked questions

### Is screenpipe free?
The source is available under the Screenpipe Commercial License, with free personal, non-commercial, educational, research, and limited evaluation use. The desktop app is available on a Free tier, with paid Basic, Business, Business Max, Business Ultra, and Enterprise subscriptions. Lifetime licenses bought before the move to subscription remain valid.

### Does screenpipe send my data to the cloud?
It depends on the deployment you choose. In local-only mode, screen captures, audio transcriptions, and extracted text are stored on your device. Data movement depends on whether you enable sync, cloud AI, exports, connectors, or team workflows. If you use a cloud AI model (like ChatGPT or Claude), the queries and context you send go to that provider. Local models via Ollama keep inference on the device.

### How much disk space does screenpipe use?
Approximately 5-10 GB per month of continuous recording, depending on screen activity. You can configure retention periods to automatically delete old data.

### Does screenpipe slow down my computer?
screenpipe is designed to be lightweight. Typical CPU usage is 5-15% on modern hardware. You can adjust capture intervals and OCR frequency to reduce resource usage.

### Can I use screenpipe with ChatGPT/Claude/Cursor?
Yes. screenpipe runs as an MCP (Model Context Protocol) server, which allows Claude Desktop, Cursor, and other AI assistants to directly query your screen history. For ChatGPT, you can use the screenpipe API to provide context.

### Is screenpipe GDPR compliant?
In local-only mode, screen captures, audio transcriptions, extracted text, and local-model inference stay on your device. Data movement depends on whether you enable sync, hosted or third-party AI, exports, connectors, or team workflows. Additional compliance documentation is available for enterprise deployments.

### What OCR engines does screenpipe support?
- macOS: Apple Vision (native, best quality), Tesseract
- Windows: Windows native OCR, Tesseract
- Linux: Tesseract

### Can I record multiple monitors?
Yes. screenpipe captures all connected monitors simultaneously. This is a key advantage over alternatives like Rewind/Limitless which only capture the active window.

### How do I uninstall screenpipe?
Delete the application and its data folder. On macOS: ~/Library/Application Support/screenpipe. On Windows: %APPDATA%/screenpipe. All your data is contained in these folders.

### Is there an iOS/Android app?
Not currently. screenpipe is a desktop application for macOS, Windows, and Linux.