| What it does | On-device voice AI SDK with no cloud dependency
| Cut Claude Code, Codex & Gemini quota burn by 60–72% with persistent AI coding memory and deterministic verification
|
|---|
| Stage | Not given
| Not given
|
|---|
| Founded | Not given
| Not given
|
|---|
| Based in | CY
| RO
|
|---|
| Pricing model | Freemium
| Freemium
|
|---|
| Pricing | Published models free for commercial use; custom models available as paid engagements.
| Explorer: free with 50 MCP calls; Oxyo Core: $19/mo (or $15/mo annually) with ~1.9M tokens/week and 300 MCP calls/week; Oxyo Pro: $49/mo (or $39/mo annually) with ~3.8M tokens/week and 1,250 MCP calls/week; Oxyo Max: $99/mo (or $79/mo annually) with ~7.5M tokens/week and 5,000 MCP calls/week. All pl
|
|---|
| Key features | - Custom wake word detection with configurable threshold
- Streaming and batch automatic speech recognition
- Custom keyword spotting for voice commands
- Voice activity detection to identify speech
- Speech-to-intent with end-to-end processing
- Production binaries under 1 megabyte
- Models encrypted at rest with offline-first design
- NEON-accelerated battery-aware kernels
- Domain-specific model tuning
| - Lifelong persistent memory across sessions retained in an isolated vault
- Automatic belief revision to eliminate contradicted specifications
- Deterministic Python sandbox for verified calculations
- Model Context Protocol integration for IDE connectivity
- Causal root-cause graph tracking
- Visual learning from screenshots and diagrams
- Reusable skill synthesis for algorithmic classes
- Web console, desktop app, and REST API access
- Encrypted BYOK vault for Claude, OpenAI, and OpenRouter
|
|---|
| What makes it different | - Runs entirely on device with no cloud round-trip or per-detection fees
- Published models free for commercial use with custom tuning available
- Lightweight runtime under 1 MB with predictable latency and battery efficiency
- Models designed and tuned specifically to your product domain
- Audio never leaves the device and models are encrypted at rest
| - 60–72% reduction in token quota burn measured on production 112k-token repositories
- Zero re-sending of context; only ~1k surgical tokens retrieved per turn instead of 50k–200k
- 100% contradiction-free memory with automatic outdated fact retirement
- Unified memory across web, browser, and multiple IDEs
- Model-independent brain; memory portable across AI providers
|
|---|
| Free plan | Not given
| Not given
|
|---|
| Open source | Not given
| Not given
|
|---|
| Platforms | Not given
| Not given
|
|---|
| Public API | Not given
| Not given
|
|---|
| Community votes | 0 | 0 |
|---|
| DR (Domain Rating by Ahrefs) | Not checked yet ● no change since last check | Not checked yet ● no change since last check |
|---|
| Trust Flow (Majestic) | 3 | 0 |
|---|
| Citation Flow (Majestic) | 21 | 0 |
|---|
| Referring domains (Majestic) | 17 | 0 |
|---|
VoxRT
Oxyo