Turn any text in macOS into AI-enhanced output — instantly, locally, and without leaving your workflow.
Aya is a system-level AI assistant that enables users to rewrite, translate, and summarize text across any application using a global hotkey — powered by a local multilingual model.
Learn more about Tiny Aya: https://cohere.com/blog/cohere-labs-tiny-aya
https://drive.google.com/file/d/1JuajpfD5qvZozoa-_jAfl_jq95zJB6AO/view?usp=sharing
Most AI writing tools today (e.g. Copilot, Grammarly) require:
- switching context between apps
- repetitive copy/paste workflows
- sending sensitive data to cloud APIs
Aya explores a different interaction model:
AI as a system-layer capability — not just another app.
This is especially important in privacy-sensitive environments (e.g. B2B, enterprise workflows) where sending internal content to external APIs is not acceptable.
Aya leverages Tiny Aya, a multilingual model developed by Cohere Labs.
Tiny Aya supports 70+ languages, including many low-resource languages, enabling:
- Cross-language communication
- Multilingual rewriting
- Translation without external APIs
This makes Aya particularly valuable for:
- global teams
- international operations
- multilingual product environments
Aya is built as a local-first AI system, which directly addresses enterprise concerns:
- No data leaves the device
- No third-party API calls
- Suitable for internal documents, Slack messages, Notion content
- No per-token API cost
- Predictable infrastructure usage
- Works offline
- No dependency on external services
This design aligns with how many B2B teams evaluate AI adoption:
control, compliance, and predictability over convenience.
- Fix grammar instantly
- Translate (multilingual, 70+ languages)
- Rewrite professionally
- Summarize text
- Global hotkey (
Cmd + Shift + .) - Works across any app (Notion, Slack, browser, etc.)
- Fully local inference (Tiny Aya via aya-cli)
- User selects text in any application
- Aya captures the selection via clipboard
- Sends request to local AI model (
aya-cli) - Receives structured output
- Replaces or returns improved text
User Selection ↓ Clipboard Capture ↓ macOS Menu Bar App (Swift) ↓ Local API (aya-cli) ↓ Tiny Aya (llama.cpp backend)
- Swift (macOS menu bar app)
- aya-cli (local LLM runtime)
- llama.cpp (model inference engine)
- Tiny Aya (multilingual model)
https://github.com/Cohere-Labs/aya-cli
aya-cli serve
see demo
Clipboard-based interaction (paste simulation) No UI for mode switching yet Requires local model setup
Direct inline replacement (no paste simulation) Mode selection UI (rewrite / translate / summarize) Prompt customization Memory / personalization layer Performance optimization for real-time usage Packaging as installable macOS app
This project uses Tiny Aya via aya-cli by Cohere Labs: https://cohere.com/blog/cohere-labs-tiny-aya https://github.com/Cohere-Labs/aya-cli