Conduck next to nine alternatives, judged on the client, not the model.
Check: the maker documents it. Cross: a first-party source rules it out. Question mark: unknown, the sources are silent. Open a row for the reasons.
Pick the app you use today.
| Capability | Conduck Apple-native agent client | AskWatch Apple AI client | OpenClient Open-source Apple client | Conduit Self-hosted mobile client | OpenCat Multi-provider Apple client | Open WebUI Self-hosted workspace | Chatbox AI Cross-platform client | ChatGPT Managed assistant | Noema Local-AI workspace | BoltAI Mac AI workspace |
|---|---|---|---|---|---|---|---|---|---|---|
| Apple surfaces | ||||||||||
| Yes: Native iPhone app with Action Button, Control Center, Shortcuts, share sheet and Safari webpage sharing, voice, text, photos, and files. Native iPad interface with the same conversations and connections. | Yes: Full chat and voice app, widgets, Control Center, Shortcuts, and Action Button support. Full app on iPad; Watch settings remain an iPhone responsibility. | Yes: Native iPhone chat app with streaming, attachments, voice, widgets, Control Center, Shortcuts, Share Sheet, and Live Activities. Native iPad app with the same chat workspace, drag and drop, widgets, Control Center, and Shortcuts. | Yes: Installed iPhone app with streaming chat, voice calls, uploads, widgets, Shortcuts, quick actions, and Share Sheet intake. Installed iPad app with an adaptive layout and the same mobile chat, workspace, voice, widgets, and sharing features. | Yes: Native iPhone app with chat, voice, local and remote models, provider tools, AI Keyboard, translation, and Siri Shortcuts. Native iPad app with the same broad model-client proposition and optional iCloud synchronization. | Limited: The main product installs as a Safari PWA on iPhone; no separate native iPhone client is documented by the project. The browser workspace and Safari PWA run on iPad; no dedicated native iPad target is documented. | Yes: Installed iPhone app for chat, models, files, images, knowledge, and optional hosted service. Installed iPad app; the App Store also lists Apple-silicon Mac and Vision compatibility for the mobile build. | Yes: Native iPhone app plus Apple Intelligence integration for Siri, Writing Tools, Visual Intelligence, Image Playground, and Shortcuts on supported systems. Native iPad app with account-synchronized chats, voice, files, projects, tools, and the supported Apple Intelligence integrations. | Yes: Native iPhone app with local and remote models, documents, tools, optional sync, voice, and model discovery on eligible hardware. Native iPad app with the same local-AI and project proposition, plus optional access to models running on an awake Mac. | Limited: BoltAI Mobile is documented as beta with account-linked access. Exact shipping status, feature parity, and public storefront availability are not established. Official account material names iPad among beta account-based platforms, but current documentation does not establish parity with the Mac app. | |
| ||||||||||
| Yes: Native Mac app, menu-bar companion, global shortcut, and Screenshot & Ask. | Yes: Native Mac app with floating panel, menu bar, Safari extension, global shortcuts, widgets, and command-line tool. | Yes: Dedicated SwiftUI Mac target with a menu-bar companion, desktop widgets, drag and drop, Shortcuts, and a New Chat system control. | Limited: The App Store offers the iPad-compatible app on Apple-silicon Macs. Public release 4.1.3 documents Mac desktop UX, but the US store version checked was still 4.1.2; no dedicated macOS target is documented. | Yes: Native Mac app with quick access, an Accessibility-based keyboard extension, slash commands, and system-wide translation workflows. | Limited: Browser or desktop PWA, plus a separate early-alpha native desktop app with quick bar, screenshots, push-to-talk, local inference, and multiple connections. | Yes: Dedicated desktop download for macOS alongside Windows and Linux clients. | Yes: OpenAI’s current Mac app combines Chat, Work, and Codex. Earlier Option-Space, screenshot, Handoff, and Work with Apps capabilities are documented for the previous app, now called ChatGPT Classic; feature parity should not be assumed. | Yes: Native Apple-silicon Mac app that can run local models and serve them to other Noema devices through Constellation. | Yes: Shipping native SwiftUI and AppKit Mac app with menu-bar access, global shortcuts, Instant Chat, inline workflows, dictation, screenshots, tools, documents, and projects. | |
| ||||||||||
| Yes: Native Watch app for local voice and text asks, direct gateway dispatch, conversations, spoken replies, Control Center, Smart Stack, and Watch Action Button. | Yes: Native Watch app with voice, complications, double-tap, Ultra Action Button, and documented use away from the iPhone after settings are cached. | No: The App Store lists iPhone, iPad, and Mac only, and the public project has no Watch target. | No: The README documents iOS and Android only, and the public project has no Watch target. | No: The App Store compatibility list names iPhone, iPad, Mac, and Apple Vision only. | Unknown: No Apple Watch app or Watch-specific surface is documented in the official project sources checked. | No: The platform list names Windows, macOS, Linux, iOS, Android, and web; no Watch app. | No: The App Store compatibility list names iPhone, iPad, and Apple Vision only; no watchOS app. | No: The maker lists iPhone, iPad, Mac, and Vision Pro; no Watch app. | No: The maker lists Mac, iPhone, and iPad; no Watch app. | |
| ||||||||||
| Yes: Official voice-first CarPlay app with recent conversations, multi-turn listening, gateway choice, mute, and end controls. | Yes: Voice-first CarPlay experience included with the iPhone app. | No: The public project has no CarPlay scene or entitlement, and the App Store listing does not mention CarPlay. | Limited: Public source contains CarPlay scene and entitlement scaffolding, but the official product and store pages checked do not establish a shipping CarPlay experience. | Unknown: No CarPlay app or CarPlay-specific experience is documented in the official sources checked. | Unknown: No CarPlay app or CarPlay-specific experience is documented in the official project sources checked. | Unknown: No CarPlay app or CarPlay-specific experience is documented in the official sources checked. | Yes: Official CarPlay experience on supported iPhone software for starting, resuming, and speaking in eligible chats; it does not control the vehicle or use vehicle and map data. | Unknown: No CarPlay target or CarPlay-specific experience is documented in the current official sources checked. | Unknown: No CarPlay app or CarPlay-specific integration is documented in the official product and help material checked. | |
| ||||||||||
| System entry points | ||||||||||
| Yes: An App Shortcut starts a turn from the Action Button, Control Center, or the Shortcuts app. No Siri voice trigger. | Yes: Siri Shortcuts are documented. | Yes: App Shortcuts and Siri, plus Spotlight indexing and custom URL links. | Yes: iOS App Intents and Shortcuts, plus widgets and quick actions. | Yes: Siri Shortcuts are documented. | Unknown: Apple App Intents, Action Button, and Control Center are not documented. | Unknown: No official Shortcuts, App Intents, or share-sheet matrix was established in the sources checked. | Yes: Apple Intelligence integration with Siri, Writing Tools, and Shortcuts. | Yes: Siri, App Intents and Shortcuts for named actions, plus Spotlight exposure. | Unknown: Shortcuts or App Intents are not documented; the client is Mac-first with mobile in beta. | |
| ||||||||||
| Yes: Menu-bar companion, configurable global hotkey, and Screenshot & Ask on the Mac. | Yes: Mac menu bar, global shortcuts, floating panel, and a command-line tool. | Yes: Mac menu-bar companion with desktop widgets and a New Chat system control. | No: No Mac app: the README documents iOS and Android only. The iPad build can run on Apple-silicon Macs. | Limited: Mac quick access, an AI Keyboard, and slash commands are documented; a menu-bar item specifically is not. | Limited: The early-alpha desktop app has a Shift-Command-I floating bar, screenshot capture, and push-to-talk. | Unknown: A macOS desktop download exists; menu-bar or hotkey access is not documented. | Limited: Option-Space and screenshots are documented for ChatGPT Classic; parity in the current Mac app should not be assumed. | Unknown: Native Mac app; menu-bar or hotkey access is not documented. | Yes: Menu bar, global shortcuts, Instant Chat, and inline commands over selected text. | |
| ||||||||||
| AI connections | ||||||||||
| Yes: OpenRouter with your own key, no server needed. | Yes: Many named provider APIs and OpenRouter with your own key. | Limited: One configured base URL to a LiteLLM, Ollama, or compatible OpenAI-style server; hosted providers go through that server. | Yes: First-party OpenRouter plus direct Chat Completions or Responses endpoints. | Yes: Direct API-key connections to named providers and custom OpenAI-compatible services. | Yes: OpenAI, Anthropic-compatible, and OpenAI-compatible routes with your keys, configured on the server. | Yes: OpenAI, Anthropic, Gemini, OpenRouter, and custom OpenAI-compatible endpoints. | No: Not a BYO-key client; it connects to OpenAI’s managed service and account. | Yes: OpenRouter, OpenAI-compatible endpoints, and custom endpoints. | Yes: Direct BYO providers, OpenRouter, custom endpoints, CLI subscriptions, and hosted BoltAI models. | |
| ||||||||||
| Yes: Ollama, LiteLLM, Open WebUI, LM Studio, vLLM, or any compatible OpenAI-style endpoint. | Yes: Ollama, vLLM, llama.cpp, and other OpenAI-compatible servers. | Yes: LiteLLM, Ollama, or a compatible OpenAI-style base URL. | Yes: Open WebUI, native Ollama, LM Studio, and compatible endpoints. | Yes: Custom OpenAI-compatible services and a self-hosted Team service. | Yes: It is one: Ollama, local servers, custom pipes, and extensive provider options. | Yes: Ollama, LM Studio, and custom OpenAI-compatible endpoints. | No: Not a client for arbitrary OpenAI-compatible endpoints. | Yes: LM Studio, Ollama, and a Mac serving models to other Noema devices. | Yes: Ollama, LM Studio, and custom OpenAI-compatible endpoints. | |
| ||||||||||
| No: Conduck is a client. Models run on your server or provider, never on the phone. | Unknown: Local models run on Ollama, vLLM, or llama.cpp servers you host; in-app inference is not documented. | No: The README routes local models through a running Ollama server; the app itself contains no inference engine. | Yes: Apple’s offline on-device model on eligible iOS 26 devices with Apple Intelligence. | Yes: Local MLX and Apple models. | Limited: The desktop app on Mac can run models through a built-in llama.cpp engine; on iPhone, models run on the server. | Unknown: On-device inference is not documented; local models run through Ollama or LM Studio. | Unknown: On-device inference is not documented; Apple Intelligence integration routes to OpenAI. | Yes: Several on-device runtimes and Apple Foundation Models. | Unknown: External local-model servers are documented; in-app inference is not. | |
| ||||||||||
| In the conversation | ||||||||||
| Yes: Apple on-device speech recognition and read-aloud by default. Optional BYO-key cloud voice through OpenAI, ElevenLabs, Gemini, Mistral, or OpenRouter. | Yes: Dictation, push-to-talk, turn-based voice, and low-latency live conversation, varying by mode and surface. | Yes: Apple Speech or server-side transcription; documented read-aloud sends reply text to the configured server. | Yes: On-device or server speech recognition, plus a hands-free voice-call mode. | Yes: Voice chat, voice input, read-aloud, and several provider speech paths. | Yes: Voice and video conversations with configurable speech engines, plus desktop push-to-talk. | Yes: Voice features are listed; the sources checked do not detail defaults or surfaces. | Yes: Conversational voice with live interruption and background conversation; features vary by plan, device, and region. | Yes: On-device Whisper voice input and a hands-free voice beta. | Yes: Push-to-talk, Instant Dictation, Apple or cloud speech, and read-aloud; a continuous voice loop is not documented. | |
| ||||||||||
| Yes: Whatever your gateway runs: OpenClaw or Hermes brings its tools, MCP, memory, and long-running work. A plain model endpoint is chat only. | Yes: Built-in web search, Home Assistant control, weather and finance context, and prompt templates. MCP is not documented. | Yes: A client-side agent loop over LiteLLM web search and MCP tools, with per-tool approval. | Yes: Open WebUI tools, search, and skills, plus Hermes tool work that asks before sensitive steps. | Yes: MCP and tool calls, provider web search, image generation, and uploads. | Yes: Server-side tools, web search, MCP, OpenAPI, Python tools, code execution, and automations. | Yes: Web search, artifacts, local knowledge bases, and MCP. | Yes: Search, deep research, data analysis, connected apps, and custom GPTs, all inside the managed service. | Yes: Remote MCP, sandboxed Python, web and local retrieval, calendar and device tools. | Yes: Local and remote MCP, scoped filesystem access, web search, code execution, and workflows. | |
| ||||||||||
| Yes: Sends complete files to your agent and brings back the files it creates through your own file server, one output folder per dispatch. Work keeps materials and returned files together in projects. | Yes: Images, camera and photo input, and shared context. | Yes: Images and PDFs join prompts, generated images come back, and media is organised in the client. | Yes: Uploads, Open WebUI files, server files, and workspace attachments; exact behaviour depends on the connected server. | Yes: PDF and text upload, image input and generation. | Yes: Rich upload, server files, knowledge collections, generated assets, and code. | Yes: Documents, images, local knowledge, artifacts, and generated content. | Yes: Files, project sources, images, data-analysis outputs, and generated images inside the managed service. | Yes: Documents, books, images, datasets, Knowledge Packs, and generated charts, all kept local. | Yes: Documents, persistent attachments, project knowledge, images, and scoped filesystem MCP. | |
| ||||||||||
| Your data | ||||||||||
| Yes: One thread on your devices, synced through your private iCloud across iPhone, iPad, Mac, and Watch. | Limited: Keys sync through iCloud Keychain and Handoff is documented; full history sync across every surface is not established. | Yes: Optional iCloud file sync covers conversations, attachments, profile, memory, and templates. | Limited: Local-first; new Direct chats can sync into an Open WebUI account. No iCloud history model is documented. | Limited: Pro adds iCloud sync for conversations and settings; the storage and encryption model is not published. | Yes: Chats live in the instance database and follow you across browsers and devices. | Unknown: Local-first; no official iCloud or automatic cross-device history sync claim was found. | Yes: Chats, projects, memories, and files synchronize through OpenAI’s service. | Yes: Optional private-iCloud sync, off by default. | Limited: Optional BoltAI Cloud Sync on BoltAI-operated Supabase infrastructure. | |
| ||||||||||
| Yes: Requests go from your device straight to the AI you configured. No Conduck-operated route. | Limited: BYO text goes direct; managed AI uses an AskWatch proxy and default voice can pass through its Cloudflare Worker. | Yes: No OpenClient-hosted chat relay; requests go to the configured server. | Yes: No maintainer-operated data relay; configured services receive what you send. | Yes: BYO requests go direct; OpenCat Cloud is an optional publisher-operated route. | Yes: Self-hosted with no external calls by default; the instance operator is part of the trust boundary. | Yes: Local-first BYO use sends prompts to the configured provider; hosted plans use the Chatbox service. | No: OpenAI operates the service; consumer content may be used for model improvement unless disabled. | Yes: Ordinary local use needs no account; providers and optional services add network paths. | Yes: Direct BYO traffic goes to the provider; hosted models and sync use BoltAI infrastructure. | |
| ||||||||||
| Yes: No account, analytics, ads, tracking, or app telemetry. | No: Firebase Analytics records anonymous usage. | Yes: No advertising or analytics SDK; a Firebase-hosted remote-configuration fetch is documented. | Yes: Privacy policy and App Store label: no third-party analytics or advertising SDKs, data not collected. | No: Collects anonymous product-interaction and crash diagnostics. | Yes: Self-hosted with no external calls by default. | No: The App Store label says data may be used to track you across other companies’ apps; the privacy policy names Google Analytics and Sentry. | No: The App Store label lists Usage Data, Diagnostics, Identifiers, and more as linked to you, for analytics and third-party advertising among other purposes. | Limited: App Store says Data Not Collected; the policy discloses limited install and ad-campaign attribution signals. | Limited: Sentry and anonymous diagnostics are documented and can be disabled. | |
| ||||||||||
| Yes: Content-free Usage ledger of attempts, reliability, response time, and reported tokens per connection, device, and model - on your device and private iCloud, never sent to us. | Unknown: Firebase Analytics and managed-plan usage figures are recorded; a private on-device reliability and token ledger is not documented. | Limited: Usage estimates and model metadata are present; a private cross-device ledger of attempts, reliability, and latency is not documented. | No: The privacy policy describes only transient in-memory error logs; there is no usage or reliability ledger. | Unknown: Anonymous product-interaction and crash diagnostics are collected; a private reliability and latency dashboard is not documented. | Limited: Administrators can observe the instance and its users; that is server administration, not a private client ledger. | Unknown: Google Analytics and Sentry are named; a publisher-blind on-device dashboard is not documented. | Unknown: Plan limits and task state are shown; a publisher-blind local ledger is not documented. | Unknown: Model fit, routing, and runtime controls are shown; a content-free reliability and latency dashboard is not documented. | Unknown: Optional Sentry reports and anonymous diagnostics are documented; a private ledger of attempts, reliability, and latency is not. | |
| ||||||||||
| Source and price | ||||||||||
| Yes: Apache-2.0 application source with the exact App Store commit and an SBOM. | Unknown: No public source repository was found in the official pages checked. | Yes: Complete public Swift project under GNU AGPL 3.0. | Yes: Developed in public on GitHub under GNU GPL 3.0. | Unknown: No official public source repository or app-source license was identified. | Yes: Public source under the Open WebUI License, which its docs say is not OSI-approved; older core is BSD-3-Clause. | Limited: GPLv3 Community Edition; not every shipping feature is established to be in public source. | Unknown: No official public source repository for the shipping clients was identified. | Limited: Apache-2.0 Community Edition that excludes the production App Store tree and proprietary components. | No: Proprietary application under BoltAI’s EULA. | |
| ||||||||||
| Yes: Free to download and use, including professional use. Optional Conduck Pro subscription lifts the free plan's three-project and three-gateway limits. | Limited: Free download with a limited BYO tier; one-time PRO purchases and managed plans. | Yes: Free with no required subscription or feature paywall; optional support purchases. | Limited: The site says no paid tiers; the US App Store listing checked showed a $3.99 price. The sources conflict. | Limited: Free download with a one-time Pro unlock and optional paid Cloud AI. | Yes: Standard branded self-hosting is free; white-labeling needs an enterprise license. | Yes: Free BYO client lane; optional hosted plans. | Limited: A free plan that may show ads; models, voice, and tools vary by paid plan. | Yes: Free consumer app with no required subscription. | No: Paid perpetual Mac license from $79. | |
| ||||||||||
We make Conduck, so this is not an independent review. Every competitor fact comes from its own product pages, store listing, privacy policy, or public source on the date shown, not from hands-on testing.
Conduck is free for individual use. Start with an OpenRouter key, or connect a self-hosted agent or compatible model server.
