Ollama Local AI

Ollama Local AI – Run Local LLMs & OpenAI Proxy for Android

Turn your Android device into a private, high-performance offline AI server and local-network OpenAI-compatible API gateway. Run open-source Large Language Models (LLMs) on-device with zero internet dependency, or transform your phone into an air-gapped, zero-cloud AI router bridging desktop IDEs—including Cursor, VS Code, and Windsurf—directly across your local Wi-Fi.

🔒 Why Ollama Local AI?

100% Offline & Air-Gapped: Run quantized GGUF models directly on device CPU/GPU via embedded llama.cpp. No telemetry, no external logs, and no cloud relays.

Local LAN Privacy: Never share API keys online. Provider configurations and credentials remain stored strictly on-device, protected with hardware-backed AES-256 encryption.

Zero Third-Party Relays: When routing to custom cloud endpoints, traffic travels directly from your device to the target provider—never through an intermediary middleman server.

Universal OpenAI API: Exposes standard endpoints (/v1/chat/completions, /v1/models, /health) to any machine on your private Wi-Fi/LAN subnet.

Workstation Resource Offloading: Route desktop coding agents and IDE extensions to your phone to preserve RAM and GPU performance on your primary development PC.

Multi-Provider Gateway: Switch seamlessly between local on-device models, self-hosted Ollama instances, llama.cpp servers, and direct cloud APIs (OpenAI, Anthropic Claude, NVIDIA NIM, Hugging Face).

🛠️ Quick Start Guide: How to Connect Your IDE via LAN

Launch Server: Open the app, select your local GGUF model (or direct cloud keys), and tap "Start Server".

Note Endpoint: Copy the displayed local network IP and port (e.g., http://192.168.1.50:8080/v1).

Configure Your IDE / Coding Assistant:
• Cursor: Go to Settings > Models > OpenAI API Key. Enter any placeholder (e.g., "local-lan"), enable "Override OpenAI Base URL", and paste your phone's endpoint URL.
• VS Code (Continue / Cline / Roo Code): In your provider configuration, set type to "openai" and configure "apiBase" or "baseUrl" to your phone's local network IP.
• Windsurf: Point custom OpenAI model endpoints directly to your Android LAN address.

Start Coding: Requests, completions, and autonomous agent loops now execute securely through your Android device in real time.

⚡ Key Features

On-Device Inference Engine: Native GGUF model execution (Llama 3, Mistral, Phi, Gemma, Qwen) powered by optimized llama.cpp acceleration.

Smart LAN Proxy & Failover: Built-in request pools, provider timeouts, rate-limiting, automatic failover, and round-robin traffic distribution.

WebX Live Canvas: Generate, test, and preview interactive frontend code (HTML, Tailwind CSS, JavaScript) directly inside chat.

Cross-Device WebUI: Access the full chat interface and server management console from any PC or tablet browser on the same network.

LAN Master/Worker Clustering: Pair multiple Android devices across your local network to pool memory and compute resources into an inference cluster.

Traffic Observatory: Monitor token generation speed, latency metrics, and real-time request payloads with built-in proxy diagnostics.

Hardware-Backed Security: EncryptedSharedPreferences (AES-256) ensures API credentials never leave local hardware unencrypted.

🔌 Supported Tools & Integrations

Connect any client supporting standard OpenAI-compatible REST endpoints:

IDEs: Cursor, VS Code, Windsurf, JetBrains, Antigravity

Agent Frameworks: LangChain, LlamaIndex, LiteLLM, AutoGen, CrewAI

Extensions & Tools: Continue.dev, Cline, Roo Code, Aider, Open WebUI, Python scripts, and cURL

Disclaimer: Ollama Local AI is an independent developer utility and local router. It is not affiliated with, sponsored by, or endorsed by Ollama, OpenAI, Anthropic, or any mentioned third-party providers.

Recent changes:
WHAT'S NEW
- Type while AI works: messages queue
- Build: "Send now" adds to running task
- Type / to pick MCP server in Build
- Ask for MCP servers in plain words
- Better preview for React, Vite, Node, Python
- Images and fonts show in preview
- Smoother thinking box
- Retry/Continue/Regenerate keep skills and rules
- More reliable saving of rules, skills, memories
- Easier MCP setup, clearer tool names
- Bug fixes