Speak. Build. Ship.
Speak to Claude Code. Ship faster.
Voice commands → ÆtherLight Claude → Code execution. Real sprint management with dependency tracking. Multi-target routing to any AI.
No credit card required
Position your cursor where you want the output, then speak your command
Writing AI Prompts Shouldn't Be
Harder Than Writing Code
Slow & Tedious
- • 30+ seconds thinking
- • 45+ seconds typing
- • 20+ seconds refining
- • 95 seconds total per prompt
Knowledge Silos
- • Juniors struggle with prompts
- • Best practices trapped in senior engineers' heads
- • Teams repeat solved problems
- • No learning from the community
Inconsistent Results
- • 67% first-try success rate
- • 2.3 iterations average
- • Wasted tokens and time
- • Frustration with AI tools
Roadmap
What we're building
1. Voice Transcription
Press hotkey to speak. Speak naturally. Position your cursor where you want the output - ÆtherLight Voice tab, terminal, or any text field.
2. Multi-Target Routing
Send to ÆtherLight Claude, Cursor, node, or any terminal you've named. Your voice goes where your cursor is.
3. Claude Code Integration
ÆtherLight Claude wrapper receives your command and executes it through Claude Code. Real-time AI responses.
4. Sprint Management
Track progress, dependencies, time estimates, phase labels. Chain of thought prompting gives Claude better context about your project.
5. Knowledge Pools & Team Sync (Roadmap 2026)
Share patterns with your team. Build collective intelligence. Mesh network for distributed knowledge across trusted developers.
Why Lumina Is Different
| Feature | Traditional Approach | Other Tools | Lumina |
|---|---|---|---|
| Speed | Type manually (95s) | Still type (60s) | Speak (1.4s) |
| Success Rate | 67% success rate | 70% success rate | 89% target* |
| Learning | No learning | Personal history | Collective intelligence |
| Patterns | Start from scratch | Use templates | Growing library (target: 10K+) |
| Privacy | Code privacy concerns | Code sent to cloud | Privacy-first architecture |
| Integration | One AI only | Limited integrations | All AI assistants |
Powered by patent-pending hierarchical architecture - 9× faster than traditional RAG while preserving complete privacy.
Our Performance Targets
Based on benchmarks and beta validation goals
* Benchmarks based on theoretical calculations. Real-world performance during beta testing.
What's Working Right Now
These features are live and ready to use. Install the VS Code extension and start building.
ÆtherLight Claude
Voice commands directly to Claude Code. Speak → Transcribe → Execute. Your voice becomes Claude's commands.
Sprint Management
Real-time progress tracking. Dependencies, time estimates, phase labels. Chain of thought prompting for better AI context.
Multi-Target Routing
Send voice to ÆtherLight Claude, Cursor, node, or any terminal. Name your terminal tabs to route intelligently.
Open Source VS Code Extension
Privacy First
Proprietary architecture. Faster than traditional RAG. Works locally with complete privacy.
Never Lose Knowledge (Roadmap)
Every solution preserved locally. Searchable instantly. Distributed across your trusted network.
Team Gets Smarter (Roadmap)
Invite teammates, share patterns, build collective intelligence that compounds over time.
See It In Action
Real screenshots from the VS Code extension. No mockups. This is what you'll actually use.
Sprint Management Built-In
Enterprise-grade project management inside VS Code. Track progress (33/60 tasks completed), view dependencies, monitor time estimates, and organize by phases.
- Real-time completion percentages
- Task dependencies and blocking issues
- Time estimates and sprint velocity
- Phase labels for better organization
Voice → Terminal Integration
Send voice commands directly to your terminal or any AI assistant. ÆtherLight Claude receives your transcribed command and executes immediately.
- "Send to Terminal" button for instant execution
- Works with ÆtherLight Claude, Cursor, or any terminal
- Clear instructions shown in the interface
- Fast transcription powered by OpenAI Whisper
Real-Time Voice Transcription
See your voice commands transcribed in real-time. Press backtick (`), speak naturally, and watch your words appear instantly with high accuracy.
- Live transcription as you speak
- "Start Listening" button shows recording status
- Command/Transcription split view
- Works in any language (90+ supported)
Smart Multi-Target Routing
Choose where your voice command goes: ÆtherLight Claude (Claude Code wrapper), Cursor, node, or any terminal. One voice interface, infinite destinations.
- Select target from dropdown: ÆtherLight Claude, Cursor, node, etc.
- Send to Terminal or Clear with one click
- Built-in instructions for easy onboarding
- Context-aware routing based on your workflow
Simple Token-Based Pricing
Start free with 250,000 tokens. Use them or buy more when you need them.
- Generous free token allocation
- High-accuracy transcription
- Full feature access
- Tokens never expire
- No credit card required
- 1,000,000 tokens/month
- Unused tokens expire monthly
- Best value for heavy users
- Cancel anytime
- Priority support
- 1,000,000 tokens
- Requires Pro subscription
- Add-on tokens never expire
- One-time purchase
- Buy when you run low
375 tokens per minute of audio. Free trial and add-on tokens never expire. Monthly subscription tokens expire if unused.
Frequently Asked Questions
Is my code private?▼
Yes. Your code never leaves your machine. We only store anonymized pattern abstractions.
What AI assistants does Lumina work with?▼
Claude Code, Cursor, ChatGPT, Copilot, and any AI tool that accepts text input.
Do I need an internet connection?▼
Core features work offline. Pattern sharing requires internet.
What languages are supported?▼
English, Spanish, Tagalog, Cebuano, 90+ languages supported. Dialect-aware transcription.
How does the collective intelligence work?▼
Anonymous patterns shared opt-in. 5-tier hierarchy: Individual → Team → Organization → Industry → Universal.
Can I use this with my proprietary codebase?▼
Yes. Privacy-first architecture. No code uploaded. Only abstract patterns shared if you opt-in.
Join 1,000+ Engineers
Building Faster
Get early access to Lumina. Lock in beta pricing. Shape the future of AI-assisted development.
1,647 engineers on waitlist