A deep dive on everything the free, open-source app does. All of it runs on your device — no cloud inference, no account, no API key. Each feature has its own page:
| Feature | What it does |
|---|---|
| The Gateway | One local OpenAI-compatible API for every model — chat, vision, image, audio, embeddings. |
| Chat | Text + vision + reasoning, streaming, tabs, tools, voice mode, project scoping. |
| Image generation | Text→image and image→image (SDXL GGUFs), styles, LoRA, live previews. |
| Voice & speech | Speech→text (whisper) and text→speech (Kokoro), hands-free voice mode. |
| Projects (RAG) | Group chats, upload docs, chat grounded in them with cited retrieval. |
| Artifacts | HTML / React / SVG / Mermaid / Markdown rendered in a sandboxed canvas. |
| Connectors (MCP) | Add Model Context Protocol servers (none / token / OAuth), use them in chat. |
| Models | Curated, size-bucketed catalog + Hugging Face search; per-modality active model. |
| Topic | What it covers |
|---|---|
| Privacy & data | 100% local inference, encryption at rest, fully offline. |
| Architecture (open core) | The AGPL core, the pro/ submodule, and the activate() seam. |
| Off Grid Pro | The sees / remembers / acts layer — coming July 2026. |
New here? Start with the Gateway (point any OpenAI client at
http://127.0.0.1:7878/v1) or just open Chat.