Your AI. Your hardware.
Your data.
Phantasm is a fast, self-hostable AI chat app for iOS that talks to a backend you control. Point it at your own machine and chat with local models — no accounts, no cloud, no one in the middle.
Everything runs on machines you own
Private by design
Your conversations never leave your network. The app takes a URL and a token — nothing else. No account, no telemetry, no third party.
Fast, native iOS
A lean SwiftUI client built for streaming. Smooth token-by-token responses, markdown rendering, and image support that feels at home on iPhone.
Works with what you have
OpenAI-compatible at its core. Point it straight at a bare Ollama instance for plain chat, or add the orchestrator for more.
Web search built in
The optional orchestration server gives your local models live web search and page fetching — answers grounded in today, not the training cutoff.
Image generation
Hook up ComfyUI and generate images right in the chat. The server handles the pipeline; the app just shows the picture.
One app, many backends
Capabilities are detected at runtime, so the app adapts to whatever your server offers. Upgrade the backend; the app follows.
How it works
-
1
Run the backend
Spin up Ollama on any machine you own — a home server, a gaming PC, a Mac in the closet. Optionally add the Phantasm orchestrator (one Rust binary,
docker compose up) for web search and image generation. -
2
Point the app at it
Install Phantasm from the App Store, then enter your server's URL and token. That's the entire setup — the app discovers models and capabilities on its own.
-
3
Chat
Streaming responses, tool use, and generated images — all served from your own hardware. Nothing to subscribe to, nothing to leak.
Open source, end to end
The iOS app, the orchestration server, and the interface contract between them are all open. Read the code, run it yourself, make it yours.
View on GitHub