Conduck Your AI · your keys · your data
Your AI.
Every Apple device.
No middleman.
A native client for the AI gateway you host.
Conduck A native client for the AI gateway you host.
OpenClaw, Hermes, Ollama, LiteLLM - even an agent you built, with a small adapter in front
Paste a URL and key, or scan a setup code. Then talk or type from anywhere.
Paste your URL and token, or scan a code from the guided setup.
POST /v1/chat/completions
Authorization: Bearer ********No gateway yet? Start with a hosted model through OpenRouter, then move to your own when you're ready.
Type a message or tap the mic - same composer, every surface. Attach a photo, screenshot, or file when words aren't enough.
Start on Mac, finish on Watch. Your threads sync through your own iCloud.
Native on each. One conversation, synced across all of them.
Requires iOS, iPadOS, macOS, and watchOS 26.5 or later.
Every conversation lives in your own iCloud - a private CloudKit database that syncs across your devices.
Your API keys and tokens live in the Apple Keychain. With iCloud Keychain on, they sync across your own devices, end-to-end encrypted.
Your model and your files live on the AI gateway you run. The app keeps at most a small preview with the chat - in your own iCloud, never with us.
No Conduck account, no intermediary server, no telemetry. There's nothing here for us to read, leak, or hand over.
You pay for your AI. You depend on it. You tell it things - so it should answer to you, not route through someone else's servers. Conduck is a deliberately thin client: it carries your words to the gateway you run and brings the answer back. There's no intermediary service in the middle with your conversations to inspect, sell, or leak. Free for individuals, and built to stay that way.
A gateway is a server that speaks the OpenAI chat API - your own AI, reachable over HTTPS. Conduck is the client; you bring the gateway. Run OpenClaw or Hermes on an always-on Mac mini or a small VPS, or point Conduck at Ollama, LM Studio, or LiteLLM. No server yet? Start with a hosted model through OpenRouter in a few minutes, then move to your own. Once you have one, connect it by pasting its URL and token into Conduck's settings. Prefer to skip the typing? conduck-connect - our optional, open-source setup helper - exposes it over HTTPS and pairs it to the app from a single command you can read before you run.
Any server that speaks the OpenAI chat API over HTTPS. OpenClaw and Hermes work out of the box; Ollama, LM Studio, LiteLLM, or a hosted model via OpenRouter all work too. Built your own agent? Put a small adapter in front and Conduck can talk to it. Tools, file transfer, and agent memory depend on the gateway you run - a full self-hosted gateway does more than a hosted chat endpoint.
Yes. Attach photos, screenshots, and code or text files inline, or share anything into Conduck from another app. Your agent's tools can even read and write real documents through a file server you run. Vision and file features depend on what your gateway and model support.
Apple on-device transcription is the default - your voice never leaves the device. Prefer a cloud voice? Pick one and use your own key; the audio goes straight from your device to that provider, never through us. Spoken replies are off by default everywhere except CarPlay.
Yes. For a gateway you reach over a private network like Tailscale or your LAN, Conduck supports an explicit keyless mode - no token to manage. HTTPS is still required.
It can't. There's no intermediary server in the path and no telemetry - your messages go straight from your device to the gateway and providers you chose, under your own keys, over HTTPS. The App Store label reads "Data Not Collected."
No catch. Conduck is free for personal use: no paywall, no ads, no usage limits, no locked features. With no intermediary servers, there's almost nothing for us to run - so you pay only your own gateway hosting and provider keys, billed directly by those services.
For personal use, Conduck itself costs nothing. You pay for your own gateway hosting - a small VPS or an always-on Mac mini - and your own model and speech keys, billed directly by those providers. Nothing flows through us.
Your conversations sync across your iPhone, iPad, Mac, and Watch through your own iCloud - start on one device, finish on another. It's your private database; we have no access.
iPhone, iPad, Mac, Apple Watch, and CarPlay - one native client, with your conversations synced across all of them. Requires iOS, iPadOS, macOS, and watchOS 26.5 or later.
We're actively exploring it. Conduck is Apple-native today - iPhone, iPad, Mac, Apple Watch, and CarPlay - and Android isn't announced yet, but it's something we're actively looking into.
Available now for iPhone, iPad, Mac, Apple Watch, and CarPlay.
