CLI

below is the detailed cli reference for screenpipe’s core functionality.

​

core options

• fps (\-f, --fps <FLOAT>): frames per second for continuous recording
  • default: 1.0 (non-macos), 0.5 (macos)
  • storage impact: 1 FPS ≈ 30 GB/month, 5 FPS ≈ 150 GB/month
• port (\-p, --port <INT>): port to run the server on
  • default: 3030
• data-dir (--data-dir <PATH>): data directory
  • default: $HOME/.screenpipe
• debug (--debug): enable debug logging
  • default: false

​

audio options

• audio-chunk-duration (\-d, --audio-chunk-duration <INT>): audio chunk duration in seconds
  • default: 30
• disable-audio (--disable-audio): disable audio recording
  • default: false
• audio-device (\-i, --audio-device <STRING>): audio devices to use (can specify multiple)
• realtime-audio-device (\-r, --realtime-audio-device <STRING>): devices for realtime transcription
• list-audio-devices (--list-audio-devices): list available audio devices
• audio-transcription-engine (\-a, --audio-transcription-engine <ENGINE>): transcription engine
  • options:
    • deepgram: cloud-based, high quality (free tier available)
    • whisper-tiny: local, lightweight, privacy-focused
    • whisper-large: local, higher quality than tiny
    • whisper-large-v3-turbo: local, highest quality
  • default: whisper-large-v3-turbo
• enable-realtime-audio-transcription (--enable-realtime-audio-transcription): enable realtime transcription
  • default: false

​

vision options

• disable-vision (--disable-vision): disable vision recording
  • default: false
• list-monitors (--list-monitors): list available monitors
• monitor-id (\-m, --monitor-id <INT>): monitor IDs to record (can specify multiple)
• ignored-windows (--ignored-windows <STRING>): windows to ignore by title
  • example: --ignored-windows "Spotify" --ignored-windows "Chrome"
• included-windows (--included-windows <STRING>): windows to include by title
  • example: --included-windows "Code" --included-windows "Terminal"
• video-chunk-duration (--video-chunk-duration <INT>): video chunk duration in seconds
  • default: 60
• ocr-engine (\-o, --ocr-engine <ENGINE>): OCR engine selection
  • options:
    • apple-native: default for macos
    • windows-native: default for windows
    • tesseract: default for linux
    • unstructured: cloud-based (free tier available)
    • custom: configurable via SCREENPIPE_CUSTOM_OCR_CONFIG

​

custom ocr engine example

# 1. setup python environment & ocr server
python -m venv venv
source venv/bin/activate

pip install fastapi uvicorn easyocr pillow numpy

# create ocr server file
cat > app.py << EOF
from fastapi import FastAPI, HTTPException
import base64
import io
from PIL import Image
import numpy as np
import time
import easyocr

app = FastAPI()
reader = easyocr.Reader(['en', 'ch_sim'])

@app.post("/ocr")
async def read_ocr(payload: dict):
    image_b64 = payload.get("image")
    if not image_b64:
        raise HTTPException(status_code=400, detail="no image data provided")
    try:
        start = time.time()
        image_data = base64.b64decode(image_b64)
        image = Image.open(io.BytesIO(image_data))
        image_np = np.array(image)
        result = reader.readtext(image_np)
        text = "\n".join([item[1] for item in result])
        confidence = sum([item[2] for item in result]) / len(result) if result else 0.0
        print(f"ocr took {time.time() - start:.2f} seconds")
        return {
            "text": text,
            "structured_data": {},
            "confidence": confidence
        }
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"ocr error: {str(e)}")
EOF

# start the ocr server
uvicorn app:app --host 0.0.0.0 --port 8000

# 2. configure screenpipe
export SCREENPIPE_CUSTOM_OCR_CONFIG='{
  "api_url": "http://localhost:8000/ocr",
  "api_key": "",
  "timeout_ms": 5000
}'

# 3. run screenpipe with custom ocr
screenpipe add \
  path/to/videos \
  --ocr-engine custom \
  --data-dir /tmp/sp

# 4. cleanup
# stop server with ctrl+c
deactivate
rm -rf venv app.py

​

language & privacy

• language (\-l, --language <LANG>): languages to support (can specify multiple)
• use-pii-removal (--use-pii-removal): enable PII removal from OCR text
  • default: false

​

voice activity detection

• vad-engine (--vad-engine <ENGINE>): voice activity detection engine
  • options: silero, webrtc
  • default: silero
• vad-sensitivity (--vad-sensitivity <LEVEL>): VAD sensitivity level
  • options: low, medium, high
  • default: high

​

experimental features

• enable-llm (--enable-llm): enable local LLM API
  • default: false
• enable-ui-monitoring (--enable-ui-monitoring): enable UI monitoring (macos only)
  • default: false
• enable-frame-cache (--enable-frame-cache): enable experimental video frame cache
  • default: false
• capture-unfocused-windows (--capture-unfocused-windows): capture unfocused windows
  • default: false
• enable-realtime-audio-transcription (--enable-realtime-audio-transcription): enable realtime transcription
  • default: false
  • requires: at least one --realtime-audio-device
  • note: experimental feature, may impact system performance

​

subcommands

​

pipe management

# list all pipes
screenpipe pipe list [--output <FORMAT>] [--port <PORT>]

# install a new pipe
screenpipe pipe install <URL> [--output <FORMAT>] [--port <PORT>]

# get pipe info
screenpipe pipe info <ID> [--output <FORMAT>] [--port <PORT>]

# enable/disable pipe
screenpipe pipe enable <ID> [--port <PORT>]
screenpipe pipe disable <ID> [--port <PORT>]

# delete pipe
screenpipe pipe delete <ID> [-y] [--port <PORT>]

# purge all pipes
screenpipe pipe purge [-y] [--port <PORT>]

​

add external data to screenpipe (OCR only)

allows you to add external screen recordings to screenpipe, for example it could be your iphone screen recordings, your physical journal photos, etc.

companies use this to index their product’s screen recordings, for example.

# add video files
screenpipe add <PATH> [--data-dir <DIR>] [--output <FORMAT>] [--pattern <REGEX>] [--ocr-engine <ENGINE>] [--metadata-override <PATH>]

by default, screenpipe extracts metadata (fps, duration, creation time) directly from video files. however, you can override these with a metadata file:

​

metadata override example

# 1. setup test videos
mkdir -p $HOME/Downloads/tmpvideos

# copy 3 videos with random names
find $HOME/.screenpipe/data -type f -name "*monitor*.mp4" | head -n 3 | while read -r file; do
    case $RANDOM in
        [0-9][0-9][0-9][0-9])
            cp "$file" "$HOME/Downloads/tmpvideos/video1.mp4"
            ;;
        [0-9][0-9][0-9][0-9][0-9])
            cp "$file" "$HOME/Downloads/tmpvideos/video2.mp4"
            ;;
        *)
            cp "$file" "$HOME/Downloads/tmpvideos/video3.mp4"
            ;;
    esac
done

# 2. create metadata override file
cat > metadata.json << EOF
{
  "overrides": [
    {
      "file_path": "$HOME/Downloads/tmpvideos/video1.mp4",
      "metadata": {
        "creation_time": "2024-03-20T10:00:00Z",
        "fps": 30.0,
        "duration": 300.0,
        "device_name": "MacBook Pro"
      }
    },
    {
      "file_path": "$HOME/Downloads/tmpvideos/video2.mp4", 
      "metadata": {
        "creation_time": "2024-03-21T15:30:00Z",
        "fps": 24.0,
        "duration": 400.0,
        "device_name": "External Monitor"
      }
    },
    {
      "file_path": "$HOME/Downloads/tmpvideos/video3.mp4",
      "metadata": {
        "creation_time": "2024-03-22T09:15:00Z",
        "fps": 60.0,
        "duration": 200.0,
        "device_name": "iPad Screen"
      }
    }
  ]
}
EOF

# 3. run screenpipe with metadata override
screenpipe add \
  $HOME/Downloads/tmpvideos \
  --metadata-override metadata.json \
  --data-dir /tmp/sp

# 4. cleanup
rm -rf $HOME/Downloads/tmpvideos
rm metadata.json

note: if you don’t provide a metadata override file, screenpipe will automatically extract metadata from the video files. use overrides when you need to specify custom metadata or when the automatic extraction fails.

​

database

# run migrations
screenpipe migrate

​

Shell Completions

The screenpipe CLI supports generating shell completions for popular shells. Follow the steps below to enable autocompletion for your shell:

​

Bash

1. Add the following to your .bashrc profile:

eval "$(screenpipe completions bash)" 

2. Reload your profile:

source ~/.bashrc  

​

Zsh

1. Add the following to your .zshrc profile:

eval "$(screenpipe completions zsh)"

2. Reload your profile:

source ~/.zshrc  

​

Fish

1. Create \~/.config/fish/conf.d/screenpipe.fish and add this line to it:

screenpipe completions fish | source  

​

PowerShell

1. Add the following to the end of your profile file:

screenpipe completions powershell | Out-String | Invoke-Expression 

• For macOS/Linux, the profile is located at \~/.config/powershell/Microsoft.PowerShell_profile.ps1
• For Windows location is either:
  • %userprofile%\\Documents\\WindowsPowerShell\\Microsoft.PowerShell_profile.ps1 Powershell 5
  • %userprofile%\\Documents\\PowerShell\\Microsoft.PowerShell_profile.ps1 Powershell 6+
• To create the profile file you can run this in PowerShell:

if (-not (Test-Path $profile)) { New-Item $profile -Force }

• To edit your profile run this in PowerShell:

Invoke-Item $profile

1. Reload your profile:

. $PROFILE  

​

LLM links

paste these links into your Cursor chat for context:

• https://github.com/mediar-ai/screenpipe/blob/main/screenpipe-server/src/bin/screenpipe-server.rs
• https://github.com/mediar-ai/screenpipe/blob/main/screenpipe-server/src/cli.rs