Hendoyren

iOS Universel / Productivité

Hendoyren — your OpenClaw AI, native on every Apple device.

Hendoyren is a fast, private, native chat client for your own OpenClaw AI gateway. No accounts. No middle-man servers. Your conversations talk directly to the OpenClaw instance you control, over your own network — and they live on your device, not ours. Built in pure SwiftUI for iPhone, iPad, Apple Vision Pro, and Mac, it feels like a first-party Apple app, because that's the only kind of app it tries to be.

WHY HENDOYREN
• Private by design — no sign-up, no telemetry, no analytics. Nothing leaves your device except requests to your gateway.
• Your backend, your rules — connect to your own OpenClaw; you own the model, the data, and the bill.
• Native everywhere — one app, beautifully at home on iPhone, iPad, Vision Pro, and Mac.

FEATURES
• Streaming replies that arrive token-by-token
• Rich Markdown rendering — bold, lists, links, and code blocks
• Select, copy, and share any message — or export a whole conversation
• Optional spoken replies using your own ElevenLabs voice key
• Per-gateway chat history, synced privately through your iCloud
• One-scan setup with a QR code, an invitation code, or manual entry
• Trust-on-first-use certificate pinning, with a clear alert if a gateway's identity ever changes

TRY IT IN SECONDS
No OpenClaw yet? Request a demo invitation code and chat against a small hosted preview to see how it feels — then bring your own
gateway for the full experience.

WHAT YOU'LL NEED
• Your own OpenClaw instance with its HTTP endpoint enabled
• Network reachability to it (Tailscale recommended)

Hendoyren does one thing and does it well: it makes your OpenClaw feel like it's always a thumb-tap away.

Quoi de neuf dans la dernière version ?

Hendoyren 2.0 makes the app useful immediately, even before you connect your own OpenClaw gateway. It keeps the local-first idea — a native Apple chat client for AI you control — but adds a clearer first-run experience.

The biggest change is the new three-source chat selector. Under the title bar, you can choose exactly where a conversation goes: Your OpenClaw, Local AI, or the invite-only Demo path. Nothing is silently preselected, so the send button is no longer tied to a hidden or disconnected backend. You can see what you are using, switch deliberately, and understand why a message can or cannot be sent.

The headline feature is Local AI, a private on-device AI mode. This lets Hendoyren work without a server, account, gateway, or cloud AI service. On supported devices, Local AI can use Apple’s on-device Foundation Models. On other devices, it falls back to a built-in Qwen 2.5 1.5B model running through MLX. Hendoyren chooses the best engine automatically and labels it clearly, so you know whether you are using Apple Intelligence or the built-in model.

This fallback matters: Local AI is not limited to only the newest hardware. If Apple Intelligence is unavailable, disabled, or not ready, Hendoyren can still offer a working built-in model. The Qwen path includes a one-time model download of about 0.9GB, with clearer download expectations, progress, and retry behavior. There is also a setting to force the built-in model, even on devices that support Apple Intelligence.

Privacy is central to this feature. Local AI processes chat text on the device. Hendoyren does not receive your messages, does not run a hosted AI service, and includes no analytics or telemetry. The only network use specific to the built-in model is the one-time download of model weights; your prompts are not sent out as part of Local AI replies.

For OpenClaw users, version 2 adds lightweight group chat with @mentions. You can type @Name to route a message to a specific agent exposed by your OpenClaw gateway. Replies are attributed by agent, and the app can autocomplete names from the existing /v1/models probe. No new gateway endpoint is required. If you do not mention anyone, the message goes to the default agent.

Version 2 also improves the reviewer and first-time user path. The invite-code Demo gateway still exists as a fallback, but Local AI is now the recommended review path because it needs no account, code, server, or private network. A reviewer can launch the app, choose Local AI, and send a real message. This addresses the earlier problem where someone without access to your OpenClaw setup could not properly test the app.

The release continues the app’s multilingual work. Hendoyren already includes broad localization coverage, and version 2 continues expanding toward native-reviewed translations, right-to-left support, locale-aware presentation, and AI replies in the user’s language. App Store metadata and reviewer notes were refreshed around the source selector, Local AI, group chat, and privacy story.

This version is scoped to iPhone, iPad, and Mac. Vision Pro is excluded from the 2.0 submission and remains on its existing 1.1 release until the planned v3 work. The project uses per-SDK build settings and release-script safeguards so iOS and macOS move to 2.0 while visionOS stays pinned to its prior version.

There is also polish. The Welcome screen now scrolls as a full page, including its buttons, so setup is easier on smaller displays. Settings and Support now include clearer QR setup instructions. The About section no longer shows the raw bundle ID. Overall, Hendoyren 2.0 makes the app more complete and easier to understand. It still works as a private OpenClaw client, but now it also offers immediate on-device AI, clearer source selection, agent-specific @mentions, better reviewer guidance, and a more polished setup flow across iPhone, iPad, and Mac.