Take the wheel on Claude Code. When your AI agent hits a wall, Claudette pings you — drop into its real terminal from your phone, by touch or by voice, do the thing only you can, and hand control back. Same tmux session, same machine, zero context lost. In this release: • Take over and hand back — one tap into the agent's live session over SSH or Olorin Relay, then return control right where you left off. • A real shell — full terminal emulation in the agent's own tmux session, not a chat box. • tmux session persistence across backgrounding and reconnect. • Echo voice — drive it hands-free: approve, redirect, or just talk. Free with system voice; premium voice in Claudette Pro. No subscription required. A real workstation, in your pocket.

Bug fixes

Cloud Mode — Code from your phone without a laptop. Managed cloud containers spin up on demand with your GitHub repos pre-loaded. Voice-first workflow with Echo auto-activation. Echo Agent — Hands-free voice mode now wires directly into your terminal. Dictate commands, hear task summaries read back, and control sessions conversationally. Relay Hardening — Horizontal scaling with Redis-backed state, rate limiting, input validation, and proper health checks. Multi-instance ready. iCloud Pairing — Pair with Apple TV automatically via iCloud keychain. No QR codes needed. GitHub OAuth — Connect your GitHub repos with a frictionless device authorization flow. Bug fixes and stability improvements.

Major Features 1. Claudette Cloud — Managed cloud containers (fly.io) so you can use Claude Code without a laptop. Includes workspace management, cloud sessions, GitHub integration, WebSocket transport, Anthropic API key service, and auto-activating Echo voice mode for cloud sessions. 2. Echo Voice Mode — Hands-free conversational voice interface: - Push-to-talk (Walkie Talkie) and full-screen PTT modes - ElevenLabs TTS reads Claude responses aloud with voice gender picker - Intent router classifies voice commands - JSONL tail watcher for conversation mode - 100 free trial tokens, IAP purchase with auto-activate - Side-by-side voice overlays 3. Home Screen Widgets — Small, medium, and large configurable widgets with: - AppIntent sections - Profile and session details - Deep links into the app - Background refresh and reload triggers 4. Structured JSON Channel — claude -p stream-json over SSH exec channel for richer, typed streaming output 5. Connection Improvements: - Tmate transport with relay and session token auth - Paste URL connection with auto-detection (ngrok, tmate, cloudflare, generic) - Tmate network mode and onboarding wizard Fixes - Echo permanentFailure flag resets on each startListening call - Speech recognition reliability and simulator PiP guards - Data race and stderr drain fixes - Simplified Echo settings UX

Claudette Echo — Hands-Free Voice Mode Talk to Claude instead of typing. Echo reads responses aloud with natural voices powered by ElevenLabs, and listens with push-to-talk or walkie-talkie mode. Choose male or female voice. Try it free with 100 trial tokens, then unlock the full experience with Claudette Echo. Home Screen Widgets Small, medium, and large widgets show your active session status, profiles, and connection details right on your Home Screen. Picture-in-Picture Keep your Claude session visible while using other apps. Control playback from the lock screen. More Connection Options Connect via tmate relay or paste any tunnel URL (ngrok, tmate, Cloudflare). New guided onboarding walks you through setup. Structured JSON Streaming Real-time structured output from Claude with the new stream-json channel. Bug Fixes - Fixed OAuth authentication URL handling - Fixed context percentage reset on /clear - Improved audio session and speech recognition reliability

Introducing Claudette Echo — hands-free voice conversation mode. Talk to Claude and get spoken responses back, no screen or typing needed. Perfect for driving, walking, or when your hands are busy. Available as an in-app purchase ($4.99).

Voice Control Room (BETA) — hands-free AI coding from anywhere. - Conversation mode: Claude speaks a summary when done, mic auto-opens for your next instruction - Ambient narration: hear agent spawns, task completions, and errors without looking at the screen - Voice review: say "review" to hear AI-generated file-by-file change summaries - Context budget: warns before you run out of context, auto-exports session state - Agent timeline: visualize when agents ran and what tools they used - Session analytics: track cost, tokens, and duration across sessions - Diff view: see what Claude changed with inline green/red diffs - Live log stream: filterable feed of tool calls, commands, and errors -Plus: background audio (works with screen off), haptic feedback, headphone-aware narration, and a Test Background Audio button in Settings.