Version 2.0 — our biggest update yet. Sign in with the AI subscription you already have, give your personas real skills, watch them build live interactive things right in your chat, and (on Mac) hand them a whole sandboxed computer to work in. Sign in with your subscription Use the provider account you already pay for, or keep bringing your own API keys. Your keys and sign-ins talk straight to the provider — no middleman, no account to create, nothing routed through us. Skills for your personas Teach your AI reusable skills that load only when needed, compatible with the open agent skills standard. Pick skills and set a thinking budget right from a persona's card. Interactive creations in chat Your AI can now build live, interactive things inline — 3D models, charts, music, physics playgrounds and generative art. Pan, zoom and explore them, then share or export any creation as a self-contained file. A computer of its own (Mac) Give each conversation a private, disposable Linux workspace, pre-loaded with popular languages and developer tools. Let your AI run real code and experiments — safely sandboxed, then thrown away. Work folders and quick files Give a project or chat its own working folder, and type "@" to instantly find and attach files from it. Your folder follows along when you fork a conversation. More ways to work - Run your own local models on device — now with image understanding - Drop spreadsheets and video into chat - Faster, more relevant conversation search - New in-app notifications with an unread badge - Precise desktop control with zoom and modifier keys (Mac) - The latest Claude, Gemini and Kimi models Plus dozens of fixes for syncing, message queuing and overall reliability.

Voice Calls — Talk to your AI personas in real-time. Speak naturally, interrupt mid-sentence, and hear responses aloud. Works with all models, on iPhone and Mac. Projects — Organize conversations across actors into focused groups with full-text search and CRDT sync. Batch Messages — Send one message to multiple personas at once and browse their responses side by side. Structured Output — Get JSON schema-validated responses from any supported model. Auto Memory — The app now extracts noteworthy facts from your chats automatically. Summarize & Truncate — Smarter context management that summarizes history before starting fresh. OpenAI Responses API — All OpenAI models now use the modern Responses endpoint. Canvas Improvements — New browser interaction tools (scroll, hover, right-click, navigate) and headless URL loading. Plus: proactive rate-limit throttling, retry improvements, and numerous bug fixes.

**Desktop App Capture (macOS)** Control any app with AI. Capture and interact with external macOS applications through live window streaming. AI can take screenshots, click elements, and type text using Accessibility APIs. **Document Editor Canvas** Collaborative markdown editing with syntax highlighting. Edit documents while chatting with AI. Auto-save, file monitoring, and direct AI file editing. Works on iOS and macOS. **Canvas Browser** Modern floating URL bar design. AI can browse websites, click buttons, fill forms, and capture full-page screenshots. Browser context automatically syncs with chat. **Background Mode** Long-press send for background processing. Multi-turn conversations survive app suspension. Progress indicators and notifications when AI finishes responding. **Message Queue** Queue messages while AI is thinking. Queued messages auto-send when AI finishes. **New AI Models** Claude Opus 4.5, GPT 5.1, gpt-image-1.5, and Gemini 3 Pro with thought signature support. AI actors can delegate to other actors via LLM.Converse. **Parallel Tool Calls** AI executes multiple tools simultaneously for faster complex workflows. **Conversation Management** Multi-select and batch delete conversations with optimized performance. **Performance** Faster actor and conversation lookups. Lazy conversation creation. Optimized date formatting and async task management. **Bug Fixes** Fixed conversation titles, Enter key on macOS, send button animations, file I/O blocking, visionOS compatibility, and various other improvements.

SMART TIME AWARENESS Your AI now understands when time has passed between conversations. Resume old chats and the AI automatically knows it's been days or weeks since you last spoke. Better context, smarter conversations. APPLE INTELLIGENCE INTEGRATION Run AI models entirely on your device with iOS 26+ Apple Intelligence. Fast, private, no internet required. Your data never leaves your device. ON-DEVICE TRANSCRIPTION Drop audio or video files into chat for instant transcription. Works completely offline with iOS 26+ speech recognition. Your conversations, your privacy. WEB SEARCH WITH PERPLEXITY Get real-time information from the web without leaving your chat. Enable web search for actors that need current information. SMARTER MEMORY MODE Tap the brain icon to search your entire conversation history for relevant context. AI automatically finds and uses past conversations to give better answers. MCP SERVER INTEGRATION Connect external tools and services via Model Context Protocol. Sign in with GitHub and other providers. Secure authentication in your keychain. BEAUTIFUL ANIMATED BACKGROUNDS (Experimental) Watch the Tree of Life grow as you chat. Dynamic branch patterns respond to your conversation length. Enable in Settings > Extras. MODERN UI IMPROVEMENTS - Better grid layouts for conversations and actors - Improved markdown formatting (italics, bold, newlines) - Visual feedback with purple glow effects - Enhanced file attachment handling PRIVACY & SECURITY - Enhanced security-scoped resource access - Better OAuth token management - Improved keychain integration Your conversations. Your device. Your control.

What's New Widgets & Quick Access - Add EdgeTalk to your home screen with new iOS widgets - Launch conversations instantly from Control Center Better Multi-Device Sync - Improved synchronization across all your devices - More reliable conversation syncing with iCloud Performance Improvements - Faster app startup and smoother performance - Better memory management for AI conversations Enhanced Media Support - Improved audio file handling and playback - Better compatibility with various file types Stability & Privacy - Enhanced data protection and device privacy - More stable app experience with fewer crashes

Fixed many bugs and some possible crashes Updated to new Shards runtime Added many SwiftUI Shards, preparing for something very cool Enabled RealityKit on iOS as well (check memories tab) Added more models support

Added support for Grok 4 (bring your own keys only for now!) and K2 (BYOK) Few minor fixes and improvements.

Settings view refactor with reset/update functionality

Minor fixes and user experience improvements.