ToDesktop Recall Desktop SDK Plugin

A ToDesktop plugin that integrates with Recall.ai's Desktop Recording SDK to enable automatic meeting recording for Zoom, Google Meet, and Microsoft Teams.

Overview

This plugin provides a complete integration between ToDesktop and the Recall.ai Desktop Recording SDK, offering:

• Automatic meeting detection for Zoom, Google Meet, Microsoft Teams, and Slack
• Recording management with start, stop, pause, and resume functionality
• Desktop audio recording for non-meeting scenarios
• Real-time events including transcription and participant data
• Permission management for accessibility, screen capture, and microphone access
• Upload progress tracking and webhook integration
• Type-safe client library for web applications

Want a tutorial?

Check out the tutorial for a step-by-step guide on how to use the Recall desktop plugin and client SDK.

Installation & Setup

Prerequisites

1. Recall.ai Account: Sign up at recall.ai and get your API key
2. ToDesktop Builder App: Create a ToDesktop application
3. Backend Integration: Set up webhook endpoints and upload token generation

1. Install Dependencies

npm install @todesktop/client-recall

3. Add Plugin to ToDesktop Builder

1. Open ToDesktop Builder
2. Install the recall desktop sdk plugin

4. Configure Plugin Preferences

In ToDesktop Builder, configure the following preferences:

• API URL: Your Recall.ai region URL (e.g., https://us-east-1.recall.ai)
• Enable Plugin: Toggle to enable/disable recording functionality
• Request permissions on startup: Automatically request required permissions

Usage

Basic Recording Workflow

import{recallDesktop}from"@todesktop/client-recall";// Initialize the SDKawaitrecallDesktop.initSdk();// Listen for meeting detectionconststopMeetingListener=recallDesktop.addEventListener("meeting-detected",async({ window })=>{console.log("Meeting detected:",window);// Get upload token from your backendconstuploadToken=awaitgetUploadTokenFromBackend();// Start recordingconstresult=awaitrecallDesktop.startRecording(window.id,uploadToken);if(result.success){console.log("Recording started successfully");}});// Listen for recording eventsconststopStateListener=recallDesktop.addEventListener("sdk-state-change",({ sdk })=>{console.log("Recording state:",sdk.state.code);});conststopUploadListener=recallDesktop.addEventListener("upload-progress",({ progress })=>{console.log(`Upload progress: ${progress}%`);});// Handle recording completionconststopRecordingListener=recallDesktop.addEventListener("recording-ended",async({ window })=>{console.log("Recording ended for window:",window.id);});// Later, remove listeners when no longer neededstopMeetingListener();stopStateListener();stopUploadListener();stopRecordingListener();

Desktop Audio Recording

For capturing audio from applications other than supported meeting platforms:

// Prepare desktop audio recordingconst{ data }=awaitrecallDesktop.prepareDesktopAudioRecording();const{ windowId }=data;// Get upload token and start recordingconstuploadToken=awaitgetUploadTokenFromBackend();awaitrecallDesktop.startRecording(windowId,uploadToken);// Stop when doneawaitrecallDesktop.stopRecording(windowId);

Permission Management

// Check permission statusconststatus=awaitrecallDesktop.getStatus();console.log("Permissions:",status.permissions);// Request specific permissionawaitrecallDesktop.requestPermission("screen-capture");// Listen for permission changesconstremovePermissionListener=recallDesktop.addEventListener("permission-status",({ permission, status })=>{console.log(`Permission ${permission}: ${status}`);});// Remove the listener when you no longer need updatesremovePermissionListener();

Backend Integration

Demo Backend Service

A minimal Express backend lives in packages/backend for demos. It exposes:

• POST /api/create-sdk-upload – calls the Recall API and returns { id, upload_token, recording_id }
• POST /webhooks/recall – logs Recall webhook payloads for inspection
• GET /health – health check

Run it with your Recall token (replace the example value with the token for your workspace):

RECALL_API_TOKEN="c5a6aaff378e5dc5a7e28b3e2853eff832ce4bde" \
npm start --workspace=packages/backend

The server defaults to https://us-west-2.recall.ai; override via RECALL_API_BASE if you use another region. Set CORS_ORIGIN to restrict cross-origin access (defaults to *).

Creating Upload Tokens

Your backend needs to create upload tokens using the Recall.ai API:

// Example backend endpointapp.post("/api/create-upload-token",async(req,res)=>{constresponse=awaitfetch(`${RECALL_API_URL}/api/v1/sdk-upload/`,{method: "POST",headers: {Authorization: `Token ${RECALL_API_KEY}`,"Content-Type": "application/json",},body: JSON.stringify({transcript: {provider: {assembly_ai_streaming: {},},},}),});constdata=awaitresponse.json();res.json({uploadToken: data.upload_token});});

Webhook Handling

Set up webhooks to handle recording completion:

app.post("/webhooks/recall",(req,res)=>{const{ event, data }=req.body;switch(event){case"sdk_upload.complete":
console.log("Recording completed:",data.recording.id);// Process completed recordingbreak;case"sdk_upload.failed":
console.log("Recording failed:",data);// Handle failurebreak;case"sdk_upload.uploading":
console.log("Recording uploading:",data);// Track upload progressbreak;}res.status(200).send("OK");});

API Reference

Main Methods

• initSdk() - Initialize the Recall SDK
• shutdownSdk() - Shutdown the SDK and cleanup
• getStatus() - Get plugin and SDK status
• startRecording(windowId, uploadToken) - Start recording a meeting
• stopRecording(windowId) - Stop recording
• pauseRecording(windowId) - Pause recording
• resumeRecording(windowId) - Resume recording
• uploadRecording(windowId) - Compatibility no-op; recordings now stream during capture
• prepareDesktopAudioRecording() - Prepare desktop audio capture

Event Listeners

Use recallDesktop.addEventListener(eventType, callback) to subscribe. Available event types include:

• meeting-detected, meeting-updated, meeting-closed
• recording-started, recording-ended, sdk-state-change (deprecated)
• upload-progress (deprecated), realtime-event, error
• permissions-granted, permission-status
• media-capture-status, participant-capture-status, compliance-message-status, shutdown
• log, network-status

Configuration

• setConfig(config) - Update plugin configuration
• getConfig() - Get current configuration
• requestPermission(permission) - Request specific permission

Development

Available Scripts

• npm run build - Build all packages
• npm run dev - Development mode with watch
• npm run test - Run tests
• npm run typecheck - TypeScript type checking
• npm run clean - Clean build artifacts

Plugin Development

• The Electron plugin uses @recallai/desktop-sdk directly; no mock setup is required.

• Before building or type-checking the client package, the sync-sdk-types script copies the SDK's TypeScript declarations into packages/client/src/generated. This runs automatically via npm run build --workspace=@todesktop/client-recall and npm run typecheck --workspace=@todesktop/client-recall, but you can invoke it manually with:

  npm run sync-sdk-types --workspace=@todesktop/client-recall

  Changelog

  • 1.3.12
    • Updated @recallai/desktop-sdk to v2.0.26
    • Pulled in upstream microphone tracking, meeting detection, recording finalization, and Zoom, Google Meet, Teams, and Safari capture fixes
    • No wrapper API changes were required; the upstream TypeScript declarations are unchanged
  • 1.3.11
    • Updated @recallai/desktop-sdk to v2.0.24
    • Pulled in upstream realtime transcription latency, Teams Web screenshare, Zoom/Google Meet PIP capture, encoding, heartbeat, status request, participant labeling, and desktop audio recording permission fixes
  • 1.3.10
    • Updated @recallai/desktop-sdk to v2.0.22
    • Pulled in upstream Teams Web support on macOS, transcript partial-data fixes, capture reliability and performance improvements, and compliance messaging fixes
    • Added pass-through typing for the new compliance-message-status SDK event
  • 1.3.9
    • Updated @recallai/desktop-sdk to v2.0.19
    • Pulled in upstream Google Meet, Zoom, and Teams detection fixes, video/audio capture reliability fixes, encoding performance improvements, and presigned URL retry handling
    • Added pass-through support for the optional prepareDesktopAudioRecording SDK config argument
  • 1.3.8
    • Updated @recallai/desktop-sdk to v2.0.14
    • Pulled in upstream Google Meet detection fixes, macOS memory leak fixes, audio-pipeline fixes, Teams Gallery fallback capture fixes, and Teams Windows app-hang fixes
  • 1.3.7
    • Updated @recallai/desktop-sdk to v2.0.13
    • Pulled in upstream app-hang fixes, performance improvements, Google Meet PIP black-recording fixes, Arc window capture support, audio-pipeline fixes, and Google Meet detection fixes
  • 1.3.6
    • Updated @recallai/desktop-sdk to v2.0.12
    • Pulled in upstream crash fixes, TLS handling fixes, audio-pipeline telemetry, Teams/Chromium/Safari capture fixes, Zoom multiwindow sharing, and Chrome vertical-tab meeting detection
    • Reflected upstream deprecation of sdk-state-change; use recording-started and recording-ended for recording state transitions
  • 1.3.5
    • Updated @recallai/desktop-sdk to v2.0.11
    • Pulled in upstream meeting-detection, mic-stream, and audio-pipeline reliability fixes
  • 1.3.4
    • Updated @recallai/desktop-sdk to v2.0.10
    • Aligned wrapper types with new Recall events and permissions
    • Documented the streamed upload model and deprecated uploadRecording
  • 1.3.3
    • Updated @recallai/desktop-sdk to v2.0.8
  • 1.3.2
    • Updated @recallai/desktop-sdk to v2.0.4
  • 1.3.1
    • Updated @recallai/desktop-sdk to v2.0.3
  • 1.3.0
    • Updated @recallai/desktop-sdk to v2.0.0
  • 1.2.0
    • Updated @recallai/desktop-sdk to v1.3.5