Phantasm

Phantasm is a private AI chat client for the backend you already run. Point it at your own server: Ollama, vLLM, llama.cpp, or any OpenAI-compatible endpoint and chat with your own models over your own network. No account, no middleman, no data collection.

YOUR SERVER, YOUR RULES
• Connect with just a URL and a token — works against a bare Ollama out of the box
• Add the open-source Phantasm orchestrator to unlock web search, image generation, and research modes
• Multiple backend profiles for home network, VPN, or tunnel setups
• Your token lives in the iOS Keychain; conversations are stored only on your phone

A REAL CHAT EXPERIENCE
• Fast streaming responses with markdown, syntax-highlighted code blocks with copy, and charts
• Attach photos and files — vision models can see your images
• Generated images appear inline; save or share them
• Dictate messages with on-device transcription and have replies read aloud
• Long generations survive backgrounding and pick up where they left off
• Ask Phantasm from Siri and Shortcuts

TOOLS ON YOUR TERMS
• Per-chat toggles for web access and image generation
• Optional device tools — location, calendar, Apple Health — are off by default and only shared with your own backend when you enable them
• See the model's reasoning when a thinking model is selected

PRIVATE BY DESIGN
Phantasm operates no servers and collects no data. The app talks exclusively to the backend you configure — nothing else. What you say to your AI is between you and your hardware.

Phantasm requires an OpenAI-compatible server that you provide (for example, Ollama running on your own machine). Feature availability depends on your backend's capabilities.