Getting Started with Raven in Under 5 Minutes

Raven runs locally on your desktop. There's no account to create, no cloud service to sign up for, and no bot to invite to your meetings. Here's everything you need to go from zero to a fully working AI meeting copilot.

What you'll need before starting

• A Mac running macOS 12 (Monterey) or later, or a PC running Windows 10/11

• Node.js 22+ - we recommend installing via nvm. The repo includes an .nvmrc file so you can just run nvm use.

• GStreamer - the multimedia framework that powers Raven's echo cancellation

• A Deepgram API key - for real-time transcription. You can get a free key at console.deepgram.com which includes generous free credits.

• An Anthropic or OpenAI API key - for AI assistance. Get an Anthropic key at console.anthropic.com or an OpenAI key at platform.openai.com.

Step 1: Install GStreamer

GStreamer is required for Raven's echo cancellation pipeline. Installation is quick:

On macOS, open Terminal and run:

brew install gstreamer gst-plugins-base gst-plugins-good gst-plugins-bad

On Windows, download the runtime and development MSI installers from the GStreamer website (get the MSVC 1.24+ versions). Run both installers. They'll set the necessary environment variables automatically - you can verify by checking that GSTREAMER_1_0_ROOT_MSVC_X86_64 is set in your environment.

Step 2: Clone the repo and install dependencies

Open your terminal:

git clone https://github.com/Laxcorp-Research/project-raven.git

cd project-raven

npm install

The npm install step will also automatically rebuild the native better-sqlite3 module for Electron.

Step 3: Build the native components

Raven has two native components that need to be compiled: the echo cancellation addon and the platform-specific audio capture binary.

Build the echo cancellation addon:

cd src/native/aec

./build-deps.sh

npx cmake-js compile --runtime electron --runtime-version 40.4.1

cd ../../..

Build the audio capture binary:

On macOS:

cd src/native/swift/AudioCapture

swift build -c release

cd ../../../..

On Windows:

npm install -g @napi-rs/cli

cd src/native/windows

napi build --platform --release

cd (back to the mian project-raven dir)

Step 4: Launch Raven

npm run dev

Raven will open and present a 6-step onboarding flow:

Step 1: Welcome. A brief introduction. Click "Get Started."

Step 2: API Keys. Enter your Deepgram API key and your AI provider key (Anthropic or OpenAI). Choose which AI provider and model you want to use. Raven supports Claude Haiku 4.5, Claude Sonnet 4, GPT-5 Mini, GPT-5.2, and GPT-4o.

Step 3: Permissions. Raven needs three system permissions: Microphone access (to capture your voice), Screen Recording (to capture system audio via ScreenCaptureKit on macOS), and Accessibility (to enable global keyboard shortcuts). Grant each one when prompted.

Step 4: Overlay Tour. A quick walkthrough of the overlay interface — where the transcript appears, where AI responses show up, how to use quick actions, and how to move/resize the overlay.

Step 5: Keyboard Shortcuts. Learn the key shortcuts you'll use most:

• Cmd+\ - toggle the overlay on/off

• Cmd+R - start/stop recording

• Cmd+Enter - ask the AI for a suggestion

• Cmd+Arrow Keys - move the overlay

• Cmd+Shift+Up/Down - scroll the overlay

Step 6: Ready to Go. You're set. Click "Launch Raven" to open the dashboard.

Step 5: Run your first meeting

Join any meeting on Zoom, Google Meet, Microsoft Teams, or Discord. From the Raven dashboard (or using Cmd+R), start recording.

You'll immediately see the overlay appear on top of your meeting window. As the conversation happens, the live transcript populates in real time - your words labeled as "mic" and the other person's labeled as "system."

At any point, press Cmd+Enter to ask the AI for help. You can use one of the quick actions:

• Assist - get a general AI suggestion based on the conversation

• What should I say? - get a suggested response

• Follow-up - get a follow-up question to ask

• Recap - get a summary of the conversation so far

• Or type a custom question

When you're done, press Cmd+R again to stop recording. Raven will automatically save the session with the complete transcript, all AI exchanges, and a generated summary and title. You can find it in the dashboard's session history.

What to explore next

Once you're comfortable with the basics, try:

• Creating a custom mode - set up different AI personalities for different meeting types (sales, interviews, standups)

• Uploading documents - add PDFs or docs to a mode so the AI can reference them during calls

• Adjusting settings - change your AI model, switch providers, or tweak keyboard shortcuts

• Turning stealth mode off - if you want to show a colleague what Raven looks like during a call

That's it. Five minutes from clone to copilot.