ORA for macOS
- Notarised & hardened runtime
- Universal binary (arm64 + x86_64)
- Auto-update with signed manifest
- ~80 MB download
Signed, sandboxed, and ready for daily use. Auto-updates ship with changelogs you can read before they install.
Every installer is code-signed and notarised — your OS verifies it before it runs.
Mobile
The desktop stays the engine. Your phone becomes the remote: approve agent actions, monitor background runs, send prompts to your desktop workspace, receive notifications, and hand work back when you return.
System requirements
ORA itself is light. Local model requirements scale with the model you run: 0.5B feels instant on modest laptops; 70B and 250B+ need workstation-class memory and GPU.
| Use case | Minimum | Recommended |
|---|---|---|
| ORA app + cloud AI | 4 GB RAM, dual-core CPU | 8 GB+ RAM, quad-core or Apple Silicon |
| 0.5B-3B local models | 8 GB RAM | 16 GB RAM for smooth multitasking |
| 7B-9B local models | 16 GB RAM | 16-24 GB RAM, GPU helps speed |
| 12B-14B local models | 24 GB RAM | 32 GB RAM, NVMe storage, discrete GPU preferred |
| 24B-32B local models | 32 GB RAM | 48-64 GB RAM, fast NVMe, 12 GB+ VRAM preferred |
| 70B local models | 48 GB RAM with quantized builds | 64-128 GB RAM, strong GPU for usable speed |
| 250B+ heavy rigs | 128 GB+ RAM, workstation GPU | 256 GB+ RAM, multi-GPU or enterprise NIM endpoint |
| Disk + network | 500 MB app, internet for cloud providers | 2 GB+ for embeddings; model weights need extra space |
After install
No accounts. No onboarding wizards that won't end. Open and use.
Drop your API keys into the providers you already use. Keys go to the OS keychain — ORA reads them at runtime.
Point local search at one project folder to start. You can revoke it any time; embeddings stay on your disk.
Ask anything. Hit ⌘↵ to fan it out to every model. Watch them think side-by-side. Welcome to the workspace.
Roadmap
Shipping in small, stable increments. No vapourware.
Side-by-side model columns, synthesis, provider presets, and API-key routing across Claude, GPT, Gemini, Groq, OpenRouter, NVIDIA NIM & more.
Pop-out webview, folder consent, on-device embeddings, sensitive-path redaction.
Custom agents, approval center, notification inbox, system-aware agent.
On-device models on Windows & macOS — offline chat, local file search, image generation setup, no cloud prompts. Guides at heyora.in/local-ai.
Coding studio, long-form research mode and playable game builds — all inside the desktop workspace.
Updated model catalog with 106 routes, 70 free options, 11 image-generation models, original provider logos, and smart links to Local AI or provider API-key pages.
Enterprise on-premise tier, Free plan ₹0 for 3 months, 7-day Pro trial, and ORA Assist — the on-page guide on heyora.in.
Multi-step plans with dry-runs, MCP-compatible tool integration, scheduled runs, and safer approval checkpoints before any write action.
Linux AppImage, Razorpay / UPI for Indian users, GST invoices for businesses, and guided setup for OpenAI, Anthropic, Gemini, Groq, NVIDIA NIM and OpenRouter keys.
Shared federated presets, agent libraries, approved-provider lists, cost controls, and central license management for small teams.
Companion apps for approvals, remote prompts, background-run monitoring, notifications, and handoff back to the desktop workspace. Join the mobile waitlist.
Free for 3 months · 7-day Pro trial · One email for your license · No usage telemetry