All your projects in one sidebar — the sidebar now shows every project at once, with their folders and chats.
Focus on one project — choose a project at the top of the sidebar (⇧⌘P) to show only its folders and chats in that window. Other windows keep their own choice.
Go to Project or Folder — press ⌥⌘O and type to jump to any project or folder.
Show in — right-click a chat and choose Show in to jump to its project or folder.
Reorder projects — drag projects into the order you like, or use Move Up / Move Down (⌥⌘↑ / ⌥⌘↓).
Archive projects — archive a project to hide it and its chats; unarchive it any time.
Find in Chat — press ⌘F to search the whole conversation, step through matches with ⌘G / ⇧⌘G, and turn on Show Results to see every match in the sidebar.
Text Size — make text bigger or smaller everywhere with ⌘+ and ⌘−, and reset it with ⌘0 (View menu).
Buy licenses for more Macs — Settings → License → Add More Macs opens your license on vinaa.ai, where you can see the Macs using it and buy more.
OpenRouter — connect your OpenRouter account to use hundreds of models, with web search. Use Edit Models to choose which of them show up in your model list.
More providers — native support for Groq, Together AI, Fireworks AI, Cerebras and DeepInfra.
Custom connections — add any OpenAI-compatible API by its URL and key.
Ollama Cloud — Ollama and LM Studio connections now accept an optional API key, so you can use Ollama Cloud and secured local servers.
Improved
Calmer streaming — replies no longer scroll away while you read; turn following back on in Settings → Appearance. Thinking blocks remember whether you keep them expanded or collapsed.
Fixed
Fixes — failed replies retry at most twice, Quick Look shows attachment names, Open Tools Panel is disabled when there are no tools, plus layout fixes for macOS 27.
A rebuilt, more powerful agent. We rebuilt the engine behind every chat from the ground up. It handles longer, multi-step tasks, recovers on its own when something goes wrong, and works more reliably with every provider.
Clearer errors. When something goes wrong, you'll see what happened right where it happened, and you'll always know when Vinaa is still working on a reply.
Report a Problem. Send a private report from an error card or the Help menu. You preview exactly what's sent, and it never includes your conversations.
Max Tool Steps. Choose how many tool steps the agent may take in a turn (Settings → Defaults).
Smarter tool use in Auto mode. The agent searches your connected tools and turns on the ones it needs.
Faster, cheaper long chats. Prompt caching cuts the cost of long chats, and Claude clears old tool results in long runs so it stays within its context window.
Reasoning and effort. Thinking and effort settings now follow each model's real capabilities. Reasoning from local models (LM Studio, Ollama) now shows as thinking.
Code execution. With Claude, code execution keeps its workspace across steps and turns, and files it generates are saved as attachments.