What's New in 2.3.1 Security & Privacy • Encrypted thread summaries and memory proposals for stronger end-to-end protection • Tightened access controls so removed Coven members can no longer view live activity, proposals, or realtime updates Bug Fixes • Fixed an issue where accepting a Coven invite could leave your Coven list out of sync • Resolved a bug causing Personal Strix routing to fail in certain configurations • Improved reliability when switching between Covens, reducing stale or duplicate data • Fixed edge cases where deleted Covens could briefly reappear Stability • General performance and reliability improvements under the hood

Asset Library Agent-generated files are now durable workspace objects, not chat attachments. Documents, images, and data land in a native Asset Library scoped to your personal space or coven, and agents read and write them through scoped tools. Send to Agent Hand any asset to a single agent from a native sheet. The handoff opens a real thread and turn with its lineage card intact, so you can see what was sent and what came back. Provider setup that finds your credentials AICoven now detects credentials you already have — Vertex through gcloud ADC, Bedrock through your AWS shared config, local servers on their usual ports — and parses a pasted key instead of handing you a generic form. Strix offers setup in the chat where you hit the wall, not buried in settings. Tasks across surfaces Turn an assistant's task list or a recorded decision into durable tasks without leaving the thread. Home shows a live preview of each workspace. Reliability Stop is authoritative and settles the turn immediately. Streaming renders without delay on Mac. Runs recover after a disconnect instead of stranding a turn. Thread lists and workspaces refresh when they should.

AICoven 2.2.0 makes your images, files, and agents easier to trust and work with. What's new: • See it, don't chase it. Generated images and files now show up right in your conversation, no more tapping through links. • Smarter image generation. If one model can't create your image, we automatically try another so you're not stuck. • Agents that stick with it. Long agent tasks now pick up right where they left off, even if you close the app or lose connection. And when you hit stop, it stops. • Smoother downloads. Generated files and images now download reliably every time. Small fixes that add up: no more duplicate messages cluttering your chats, and better accessibility for saved files. We're just getting started. Next up, more visibility into what your agents are actually doing.

AICoven 2.1.0 makes collaborative AI work clearer, smoother, and easier to follow. What’s new: • Shared covens are more useful, with invite emails, deep-link accept flows, live thread updates, participant metadata refreshes, and better collaboration handling. • Agent work is now easier to understand with richer run status cards, approval details, system status rows, artifacts, code block affordances, and clearer progress updates while agents are working. • New workspace surfaces help turn conversations into structured work, including task/Kanban surfaces, decision logs, and improved surface navigation on iPhone and iPad. • Chat timelines are more stable and readable, with better markdown rendering, tables, images, message selection, copy behavior, deduping, and reconnect recovery. • Workspace MCP and automation improvements make AICoven’s Knowledge Hub and background workflows more reliable for connected agents. • Provider and local-agent errors are clearer, including better handling for invalid keys, no-reply failures, routing exhaustion, and local-agent relay issues. • Photo attachments now handle more Apple-friendly formats, including HEIC/HEIF. • Push notifications, foreground agent completion banners, approval actions, and app launch routing have been tightened up. • Behind the scenes: expanded observability, Sentry crash reporting foundations, safer build/deploy guardrails, and many reliability fixes. This release is about making AICoven feel less like a chat box and more like a live workspace for humans and agents working together.

Welcome to AICoven 2.0, our biggest update yet. Navigation, knowledge, and chat have been rebuilt so multi-agent work feels faster, clearer, and easier to find. NEW IN 2.0 • New icon rail navigation: Jump between Strix, covens, and Settings in one tap, with unread cues that stay visible while you work. • Coven Hub: See the important threads and activity in each coven without drilling through extra screens. • Knowledge Hub: Memory is now Knowledge Hub, with a refreshed mobile and desktop experience, Skill entries, and clearer handling for long notes. • Voice messages: Hold to record in chat. AICoven transcribes your voice so agents can work from it. • Rich chat composer: Write with formatting for bold text, lists, code, links, and more, with cleaner chat bubbles and a compact cauldron loader while agents think. • More recognizable covens: Add custom avatars and descriptions, and manage Coven Settings directly from the workspace. • Stronger agent runs: Long turns are more reliable, local Hermes routing is clearer, and timeouts now give better feedback. • More provider options: Codex OAuth sign-in joins continued BYOK support across your cloud and local model stack. IMPROVEMENTS • Smoother drafts and new threads, including chats started from Strix. • Faster workspace switching, with recently visited covens and thread lists kept warm while staying mindful of mobile memory. As always, your API keys stay yours.

Your agent just got a lot better at keeping up with you. AGENT SKILLS Teach it your way of working Drop in a Markdown file and your agent learns a repeatable workflow: how you write, how you review, how you think. That context shows up in every conversation, with cloud models or local ones, and syncs automatically across your Apple devices. INTERRUPT & REDIRECT Got a follow-up before it's done? Send it. AICoven pauses the agent mid-thought, folds your new message in, and replies once with a complete answer that addresses everything. No more waiting. Also: fixed a bug where local agents timed out before completing the tasks.

Untethered Power: Hermes & OpenClaw Integration Your data is your business. For those who demand the highest level of privacy and zero latency, we have fully integrated Hermes and OpenClaw directly into the AICoven ecosystem. You can now run these models entirely locally on your macOS machine or host them on your own virtual machines. Under the Hood Improvements Stabilised Asynchronous Workers: We've upgraded our cloud task architecture to ensure background agent turns execute flawlessly, keeping your UI responsive. Enhanced Memory Extraction: Memory generation and summarisation now run completely in the background, making your chat experience faster and more fluid than ever. Update today and experience the most capable, controllable, and private version of AICoven yet!

AICoven just got a major upgrade for builders and teams. This release introduces our new Impeccable Designer experience and expands provider flexibility with Azure provider key support. Meet Impeccable Designer: a built-in design partner for product and UX workflows Go from concept to ship-ready UI with structured design actions like shaping, critique, auditing, hardening, and polishing No extra setup required for design workflows, no terminal and no MCP server configuration Added Azure provider key support, giving teams more choice and easier alignment with enterprise AI infrastructure Improved role tool organization for a cleaner, faster setup experience Design faster. Iterate smarter. Deploy with confidence.

Did you mistype something? You can now stop the action your agent is in the middle of and give it more information or corrections. Themes are also now available. Pick a theme to set the room you work in. Pick a typography voice to set how the app speaks. Pick a text size that fits your eyes. OpenDyslexic 3 is built in for readers with dyslexia, no extra install required. The Custom slot opens your device's font picker so you can pick any face you've installed through Settings on iOS or Font Book on macOS, and use it everywhere from chat bubbles to memories to settings. Other fixes: • Lock your phone mid-chat? No problem, your agent finishes the job and pings you when it's done. • Push notifications finally land on iPhone, and the unread badge clears the moment you read a reply. • Refreshed Activity tab with a fresh look. • Plus a pile of fixes for a smoother, more dependable AICoven.

What's New in This Version: Introducing the Activity Feed: Stay on top of everything happening in your workspaces. The brand-new Activity Feed aggregates all your unread agent replies, pending memory proposals, and critical budget alerts into one streamlined view. Streamlined Memory Proposals: We have redesigned the Memory Proposals screen to function as a true inbox. Once you approve or reject a memory, it will automatically clear from your view, allowing you to focus exclusively on what needs your attention. Interface Polish: Cleaned up the layout of memory proposal cards for a better reading experience and optimised the way buttons and text are displayed. Under the Hood: Implemented background improvements to increase the reliability of push notifications and further harden data privacy and connection stability.

Polish & Housekeeping Smoothed out a minor UI state quirk on the Login screen for a cleaner codebase. Enforced standard formatting logic across our backend services to keep things looking pristine. We’re still building, so if something feels broken or you have a question, hit us up at support@aicoven.ai.

Here is the breakdown of what’s new in AICoven: Cleaning House • Fixed a bug during onboarding We’re still building, so if something feels broken or you have a question, hit us up at support@aicoven.ai.

Here is the breakdown of what’s new in AICoven: Cleaning House • Fixed a message loading issues for longer conversations • Improved markdown rendering: Better handling of headings, bullet lists, inline links, and code blocks in AI responses. • Selectable text on macOS: Message bubbles on macOS now support drag-to-select across entire paragraphs, instead of being limited to one line at a time. Code blocks remain separate with their own copy button. We’re still building, so if something feels broken or you have a question, hit us up at support@aicoven.ai.

This is a big one. We’re finally giving the agents some hands (and a better memory). Here is the breakdown of what’s new in AICoven: Connect Your Apps You can now connect your agents to the tools you actually use—like Zapier and n8n. It’s called MCP (Model Context Protocol), but all you need to know is that your agents can now "talk" to thousands of other apps. We also built a smarter brain for tool selection, so the AI only grabs the specific tool it needs for your request instead of getting overwhelmed. More Independent Agents Autonomous mode is now on by default. This means your agents can take up to 25 steps to solve a complex problem without you having to babysit them or click "continue" every five seconds. Better Memory & Less Crashing • Smarter Memory: We’re now ruthlessly trimming and compressing chat history so the AI stays sharp during long conversations without hitting a wall. • Safety First: Deleting a chat while the AI is still typing won’t crash the app anymore (I know, finally). • Rate Limits: If an AI provider is having a bad day or is over-capacity, the app will now wait and try again gracefully instead of just giving up. Cleaning House • Model Cleanup: We hid the old, broken, or "deprecated" models from the list so you only see what actually works. • Your Keys, Your Rules: Thread summaries now correctly use your own API keys. Bug Fixes & Polish • Mac Fixes: Squashed those annoying crashes on the macOS version, specifically when switching tabs or editing roles. • Chat Improvements: Text selection in chat actually behaves now, and scrolling is much smoother. • Role Editor: The screen where you create your AI personalities got a facelift and better error messages if something goes wrong. We’re still building, so if something feels broken or you have a question, hit us up at support@aicoven.ai.

AICoven 1.0.2 is here! We've been busy making AICoven better for you. Here's what's new: What's New Sign up with Google and GitHub New connected apps: Google Search Console and Firebase. Have your agents analyse your data directly from the source. Always Up-to-Date Pricing: AI model costs now sync automatically in the background, so you'll always see accurate, real-time pricing for every model — including the latest from Google and Anthropic. Refreshed Mobile Design: We've given the mobile app a visual refresh. Covens and Profile sections now feature a sleek card-based design that feels more polished and lets our signature nebula background shine through. What We Fixed Smoother Popups & Widgets: Fixed visual glitches with popup backgrounds, widget spacing, and tapping notifications. No More Freezing: Squashed a bug that could cause the app to freeze when scrolling through long lists. More Reliable Agents: Your agents are now more dependable when running tasks in the background — web searches, multi-step actions, and tool use all work more consistently.