IIMAGINE Desktop

Ultimate privacy for data that cannot leave the building.

A standalone desktop app for running open-source AI models locally. Zero network traffic. Zero external dependencies. Chat, voice, tools, and a plugin ecosystem - all on your machine. Open source, free, no account required.

Download free Open source (MIT) · macOS · Windows · Linux
desktop / private modeRunning locally
ModelLlama 3.1 70B Q4
Engineiimagine-engine (llama.cpp)
NetworkDisabled
Data sent off device0 bytes
StorageEncrypted (SQLCipher)
Privacy tiers

Three provider levels. You choose per conversation.

The desktop app serves professionals in health, finance, and legal who need a spectrum of options - not just "local or cloud."

Local (default)

Models run via the bundled engine (llama.cpp). Data never leaves the machine. Zero network traffic. Suitable for any sensitivity level.

Regional Cloud

Your own GCP, AWS, or Azure credentials. Access GPT through Azure, Claude through Bedrock, Gemini through Vertex - within your organisation's compliance-approved cloud. No data sent to IIMAGINE.

Third-party Cloud

Direct API calls to OpenAI, Anthropic, etc. via your own API keys. For non-sensitive queries where you want frontier model capability.

Why regional cloud matters: Many professionals don't know they can access frontier models through their organisation's approved cloud environment. This is a middle ground between "100% local" and "send data to a third-party server." The desktop app makes it accessible without requiring API infrastructure knowledge.

The app

A complete local AI workstation. Not a wrapper.

IIMAGINE Desktop is a standalone application - separate from the web platform. It runs independently with no account, no subscription, and no internet connection required (after downloading a model). Everything happens on your machine.

Chat

Streaming responses, conversation history, model selector. Switch between downloaded models instantly.

Model Manager

Curated GGUF catalog from HuggingFace. Hardware scanner recommends models for your machine. Download with resume, queue, integrity verification.

Voice

Local text-to-speech (Kokoro model). Voice input. Full voice chat mode. No audio leaves your machine.

Agent Runner

Agentic task decomposition. Multi-step reasoning with tool calling. Runs entirely local.

Plugin System

WordPress-style hooks (preprocess, postprocess, sidebar, settings, mention, commands). Install, build, or generate plugins with AI assistance.

MCP Client

Model Context Protocol support. Connect external tool servers for file access, web search, databases. Local Files tool built-in.

Skills

Structured markdown skills injected into system prompts. Reusable knowledge modules for any topic.

Local API

OpenAI-compatible endpoint on localhost. Other apps on your machine can use your local models (Cursor, VS Code, custom tools).

Encrypted Storage

SQLite + SQLCipher. Conversations and data encrypted at rest. Decryption requires the running app instance.

Getting started

Download. Pick a model. Start chatting.

  • Download the app (macOS, Windows, Linux)
  • Hardware scanner detects your RAM, GPU, and CPU automatically
  • Browse the model catalog - recommends models that run well on your hardware
  • Download a model (resume-capable, progress tracked, integrity verified)
  • Start chatting. No API keys. No account. No internet once the model is local.

No technical knowledge required. The app handles engine management, GPU offloading, idle timeout, and model switching. You just pick a model and talk to it.

models / recommendedYour hardware
Llama 3.1 8BRecommended
4.7 GB · fits in RAM · fast inference
Llama 3.1 70B Q4Recommended
40 GB · GPU offload · high quality
Mistral 7BCompatible
4.1 GB · fast · good for routine tasks
hardware / detectedScanned
RAM64 GB Unified (Apple Silicon)
GPUM2 Pro (19-core)
Max model size~48 GB (with offload)
Who this is for

Professionals who cannot send data to external servers.

  • Legal - Case strategy, privileged communications, settlement discussions
  • Healthcare - Patient records, clinical notes, treatment planning
  • Finance - Portfolio strategy, insider information, M&A discussions
  • HR - Termination planning, compensation analysis, investigation notes
  • Government - Classified briefings, policy drafts, security matters
  • Compliance teams who have approved specific cloud providers (Azure, AWS, GCP) but not third-party AI APIs
  • Regulated industries where data residency matters (GDPR, HIPAA, SOC 2)
  • Anyone who wants to use AI without wondering who else can read their prompts

Some conversations should never leave your desk.

Free. Open source. No account required. Download and run.

Download for free