You're in.

Download Hey Eduardo

Choose your platform below. After installing, enter your license key when prompted.

Version 1.0 · Apple Silicon only · Requires 8 GB RAM · macOS 13+

Coming soonWindows

Find your license key

Enter the email you used to purchase and we'll show your key.

Getting started

  1. Download the app above and open the .dmg file.
  2. Install — drag Hey Eduardo into your Applications folder, then open it.macOS may say the app is from an unidentified developer. Go to System Settings → Privacy & Security and click Open Anyway.
  3. Activate — enter your license key when the activation screen appears.
  4. Grant Accessibility accessGo to System Settings → Privacy & Security → Accessibility and toggle Hey Eduardo on.
    Why: This lets Eduardo read the text you've highlighted in any app, so it knows what to help you with.
  5. Grant Input MonitoringGo to System Settings → Privacy & Security → Input Monitoring and toggle Hey Eduardo on.
    Why: This lets Eduardo detect your keyboard shortcut (⌘+Shift+Space) from anywhere, even when another app is in focus.
  6. Grant Screen Recording optionalGo to System Settings → Privacy & Security → Screen Recording and toggle Hey Eduardo on.
    Why: Only needed if you want Eduardo to capture a screenshot of your screen for visual context. You can skip this and add it later.
  7. You're set — Hey Eduardo lives in your menu bar. Highlight any text and press ⌘+Shift+Space to ask Eduardo about it.

Using screen capture & OCR

Don't want to highlight text? Eduardo can read your screen directly using built-in OCR.

Attach a screenshot

Click the paperclip icon in the Eduardo window to open the screen capture tool. Draw a selection box around any area of your screen — Eduardo will extract the text automatically using Apple's built-in Vision OCR engine (no internet required).

Visual context (vision models)

If you're using a vision-capable model (Qwen3.5 9B Vision, GPT-4o, or Claude), the screenshot itself is sent to the AI — so it can see charts, diagrams, or UI that can't be captured as text. For all other models, only the OCR text is used.

Privacy note

OCR text extraction always happens 100% on-device using macOS Vision. A screenshot is only sent to a cloud API if you've added a cloud API key and you're using a vision model. Local models never send anything off-device.

Screen capture requires Screen Recording permission. If Eduardo can't take a screenshot, go to System Settings → Privacy & Security → Screen Recording and toggle Hey Eduardo on.

AI models & API keys

Hey Eduardo works out of the box with a fully local AI — no account or API key needed. You can also connect a cloud provider for more powerful models.

To change your AI provider: open Eduardo → click the settings icon in the menu bar → go to AI Model.

ProviderModelsBest forAPI key?
Local (Qwen3.5)
Free · No key
Qwen3.5 4B — ~2.7 GB · 8 GB+ RAM
Qwen3.5 9B — ~5.5 GB · 16 GB+ RAM
Qwen3.5 9B Vision beta — ~6.6 GB · 16 GB+ RAM
Privacy-first use. Everything stays on your Mac — no internet needed.None — download model in Settings
Anthropic
Pay-per-use
Claude Sonnet — recommended default
Claude Opus — most capable, slower
Claude Haiku — fastest, cheapest
Best overall quality. Claude is excellent at nuanced writing, analysis, and reasoning.console.anthropic.com
OpenAI
Pay-per-use
GPT-4o — recommended default
GPT-4o mini — faster, cheaper
o1-mini — deep reasoning
Great for code, structured output, and vision tasks.platform.openai.com
OpenRouter
Pay-per-use
Access to 100s of models including Llama, Mistral, Gemini, and moreExperimenting with different models or finding the cheapest option for your use case.openrouter.ai

How to add an API key

  1. Get your API key from the provider's website (links above). Create an account if needed.
  2. In Hey Eduardo, click the settings icon in your menu bar to open Settings.
  3. Go to AI Model and click the provider card (Anthropic, OpenAI, or OpenRouter).
  4. Paste your API key into the field and click Save Key.Your key is stored securely in your Mac's Keychain — never in plain text or sent to our servers.
  5. Click Test Connection to confirm it's working, then pick your preferred model from the dropdown.