Conduck across your Apple devices
Use Conduck on iPhone, iPad, Mac and Apple Watch, with hands-free voice access in CarPlay.
Mac adds menu-bar access, global shortcuts and Screenshot & Ask. Conversations follow through your private iCloud.
The Conduck flight plan
What works now. What's on the horizon. What we're exploring.
No dates, percentages or launch theatre. Just the present, the next practical work, and the longer view - labelled by certainty.
23 capabilities mapped
How to read this
Status is part of the promise.
The order shows confidence, not a delivery calendar. We move an item only when the evidence changes.
Available now 12
Part of the current public Apple experience. Each surface still gets its own release check.
In development 0
Being built or validated now. No release date implied.
On the horizon 1
A chosen follow-on after the active work. Not scheduled.
Exploring 10
A researched possibility, not a commitment to ship.
Shipping Available now Part of the current public Apple experience. Each surface still gets its own release check.
No Conduck account or server in the middle
Messages go straight from your device to the AI and providers you chose - never through a Conduck server.
Credentials stay in Keychain, history stays on your devices and private iCloud, and Conduck collects no product telemetry.
All your AI connections in one app
Use a self-hosted agent harness like OpenClaw or Hermes, a model server like Ollama or LiteLLM, OpenRouter, or your own OpenAI-compatible endpoint - side by side.
Choose one connection when you start a conversation. That thread stays bound to it; Conduck never silently reroutes your context elsewhere. The free plan includes three configured connections; OpenRouter never uses a slot.
Voice, on-device by default
Use Apple on-device transcription by default, or bring your own key for a supported cloud voice service.
Spoken replies can use Apple voices or a service you choose. Cloud services receive what you send them and may charge under their own terms.
Ask Conduck without opening the app
Start from App Shortcuts, the Action Button, Control Center, the share sheet, Safari or Screenshot & Ask on Mac.
Today's Siri-invokable shortcuts are available now. They are separate from the newer Siri AI ideas below.
Send images, text and code
Send photos, screenshots, text and code into a conversation, or share them from another app.
What the AI can understand depends on the server and model you choose. Shared file-server exchange is described separately below.
Conversation history across your Apple devices
Search and continue conversations across your Apple devices while Conduck - not the AI - owns the canonical history.
Conduck stores and searches the thread locally, syncs it through your private CloudKit database when content sync is on (it is on by default and can be turned off), then sends it to the AI you selected when you ask.
Connect and test your AI
Connect with a URL and token, a readable setup code or Sign in with OpenRouter, then verify reachability, authentication, model access and certificate trust.
For custom servers, conduck-connect can also check Conduck’s request and response contract. A passing connection test does not prove every tool or capability on the other end works.
Troubleshoot inside Conduck
Check your AI connections, voice, file exchange, permissions, sync and Watch health from one private troubleshooting screen.
Opening it runs local reads only. Network, billable and write tests require an explicit tap, and copied diagnostics exclude secrets, URLs, custom names and file paths.
Usage, measured on your own devices
Open Settings → Usage to see the turns you sent, how reliably they landed, how long a full answer took and whatever tokens your AI reported - broken down by connection, by device and by model.
The measuring happens on your device and the records sit beside your conversations in your own private iCloud - there is no Conduck server for them to reach. Coverage is shown beside every figure, token detail depends on what your AI reports, and no costs appear anywhere. Clearing usage history clears it on every device.
Send files to your AI—and get files back
With a self-hosted agent and your own file server, send documents and other files to your AI, then download files it creates.
Available in the current Apple app. That agent must have file tools and access to the same folder; the in-app test checks the file-server path, while conduck-connect can also check the agent’s access.
Work: collect now, brief your AI later
Collect thoughts, voice notes, screenshots, links and files on a private desk beside your chats, group them into projects, then hand one reviewed brief to the AI you choose.
Capture from Conduck, the share sheet, Shortcuts, the Mac menu bar, Apple Watch or CarPlay. Saving to Work does not contact your AI. Write the task, select materials and choose an AI connection, then review the exact message and send. With a self-hosted agent and file exchange set up, returned files join the project. With content sync on, the desk syncs through your private iCloud; very large files stay on the device that captured them. The free plan includes three active projects.
Active work In development Being built or validated now. No release date implied.
Nothing sits at this stage right now.
Direction set On the horizon A chosen follow-on after the active work. Not scheduled.
Connect to Conduck from other apps
Publish an open setup-code format and test kit so compatible AI servers can create a paste code or local QR directly - without running conduck-connect.
Conduck controls the format, documentation and reference tools. Each external app decides whether to adopt it, and a valid setup code still does not prove that its server works.
Research Exploring A researched possibility, not a commitment to ship.
Conduck for Android
After the Apple client stabilizes, take Conduck into the Google ecosystem with a native phone and tablet app built around the same wire contract.
Phone and tablet come first. Assistant entry, Quick Settings, sharing and notifications still need implementation and real-device validation. No launch date.
Conduck on Googlebook laptops
Adapt the same Android app for Googlebook laptops, with keyboard, trackpad, resizable windows and desktop-scale layouts.
Googlebook is the next generation of ChromeOS built with Android technology, not a separate Conduck codebase. Support will be claimed only after testing on shipping hardware.
Conduck for Wear OS
A companion-first wrist experience for quick voice asks, recent threads and reply notifications.
This is an exploring-stage companion concept, not a committed follow-on. If pursued, the phone would own long-running requests, and system dictation may have different privacy trade-offs.
Business mode and work-data separation
Separate company conversations and credentials from personal data, with managed setup, connection lockdown, retention and offboarding - without a Conduck content cloud.
This is demand-gated, not committed. It would move forward only after the personal client is stable and real design partners prove the need. Any controls would remain customer-owned.
Manage company deployments
Give administrators customer-controlled enrollment, policy, diagnostics, credential lifecycle and offboarding for managed deployments.
This is demand-gated, not committed. If pursued, the management plane stays customer-hosted. Logs, quotas and users remain in each server’s own admin system - not a Conduck-hosted dashboard.
Open WebUI agents
Explore access to private knowledge and tools configured in Open WebUI from Conduck’s native Apple surfaces.
Basic model requests already pass the generic wire, but that alone is not the strategic feature. The open question is whether Open WebUI can complete useful knowledge and tool work without requiring a Conduck-specific server workaround or turning Conduck into a second Open WebUI client.
Coordinate work across your AIs
Give Conduck one larger instruction and let it divide the work between suitable specialist AI connections.
Conduck would discover what each connection can do, show the plan, tasks, progress, approvals, files and sources, and support retry or cancellation. Multi-hour work would run on your own agent, not indefinitely on the phone.
Next-generation Siri and Spotlight
Find, start and follow work through Siri, Spotlight, App Intents and Apple intelligence APIs where supported.
Availability varies by device and region. Search indexing would be opt-in; longer tasks must stay visible, bounded and cancellable.
Interactive agent results
Let an agent return native buttons, forms, tables, progress, approvals, files and citations - not only chat text.
A safe schema and fallback renderer would come before any connected AI can draw arbitrary interface inside the app.
One-time, revocable pairing
Explore short-lived one-time setup claims, separate credentials for each device and individual revocation - still direct to your own AI.
The reusable v1 format would remain supported. A security-sensitive v2 should exist only after real integrations prove the need.