> [ SYSTEM: INITIALIZING POCKET_NODE_CORE ] > [ STATUS: AIRPLANE_MODE_ENABLED ] > [ ENGINE: VULKAN_DETECTION... OK ] > [ MESH: TAILSCALE_LINK_READY ] > [ LOG: LOCAL_INFERENCE_ACTIVE ]
Deep Space AI.
If it works on a plane at 30,000 feet, it works in your office. Pocket Node is the first local-first AI tool that doesn't need a permission slip from a cloud server to function. Works in deep space — or your basement.
Witness Llama-3-8B running on-device with zero bars of signal.
Every screen. Local. No cloud required.
> [ TIER_MATRIX ]: LITE VS PRO VS CLOUD WRAPPERS
POCKET NODE LITE
FREE
Your entry into sovereignty. Run the Phi-3 Mini model with 100% privacy and zero data telemetry. Perfect for proving the 'Offline' concept before committing to the stack.
- [x] Phi-3 Mini bundled — offline, instant
- [x] Zero telemetry, zero data collection
- [x] Works in full airplane mode
- [ ] Vulkan GPU acceleration — Pro only
- [ ] Full model store — Pro only
- [ ] Tailscale mesh handoff — Pro only
// Gated behind Ghost signup. APK delivered via welcome email.
[ INITIALIZE_FREE_DOWNLOAD ]POCKET NODE PRO
$19.99 ONE-TIME
Own your mobile inference engine. Unlocks Vulkan GPU acceleration for high-speed generation and the full model store (Llama-3, Mistral). No subscriptions. No Cloud Tax. Pay once, own it forever.
- [x] Everything in Lite, plus:
- [x] Local RAG — chat with private PDFs, ePub, text files. Fully offline.
- [x] API / Edge Worker Mode — expose local inference over Tailscale. Your homelab can offload tasks to your phone.
- [x] Vision + Private Voice — LLaVA image analysis + Whisper dictation. Zero uploads to Google or Apple.
- [x] Custom GGUF Loader — import any model from Hugging Face or local storage. No curation limits.
- [x] Background Execution — model stays hot in RAM. Instant response when you switch back.
- [x] Thermal & Battery Governors — cap GPU usage, set thread limits, restrict API Edge to charging-only.
> [ MESH_HANDOFF ]: YOUR PHONE IS A NODE
> [ ROUTING: Task(Deep_Reasoning) ] > [ PEER: Mac_Studio_M1_Max ] > [ RAM: 32GB_UNIFIED_AVAILABLE ] > [ STATUS: HANDOFF_COMPLETE ]
Your phone is a node, not just a client. When a task requires deep reasoning or massive context, Pocket Node routes the request to your Mac Studio M1 Max via your private Tailscale mesh — using 32GB of unified memory to power the intelligence in your pocket. The same LiteLLM routing architecture that powers Neo, Uno, and Moolah handles the handoff.
> [ USE_CASES ]: WHO THIS IS FOR
CLOUD SKEPTICS — START FREE
Not ready to build a homelab? Pocket Node Lite is your zero-friction proof-of-concept. Run Phi-3 Mini offline on hardware you already own. See local AI work before you spend a dollar.
PRIVACY OPERATORS
Lawyers, consultants, business owners handling sensitive documents. Both tiers: zero telemetry, prompts never leave the device. No shadow IT risk. No compliance exposure. Not ever.
HOMELAB BUILDERS — GO PRO
Already running Neo and Moolah? Pocket Node Pro is your mobile interface. Full model store, Vulkan speed, and Tailscale handoff turn your phone into the last node in your ZCT mesh.