Raycast Release Notes

💻 macOS Tahoe Ready

macOS Tahoe is out, bringing many user interface changes. While there are no major new features, we've focused on implementing some of these design tweaks in Raycast, such as Liquid Glass controls in AI Chat.

💎 Improvements

• Focus: Support for Comet browser

🐞 Fixes

• AI Chat: Window can now be made full-screen again
• AI Chat: Floating window will now follow the current space
• Auto-quit: Fix video conference apps terminated while audio, video or screen sharing was active (macOS Tahoe)

🎙️ Auto Transcribe with Granola, Auto Models & Bring Your Own Models

Never take meeting notes again (😱) with Auto Transcribe, powered by Granola. Your meetings can now be transcribed and usefully summarized so you can stay focused on the discussion. Auto Transcribe works together with Auto Join Meetings so you can seamlessly jump into calls while Granola captures all the important details in the background.

Go to Settings → Extensions → My Schedule to enable and customize. You can decide which meetings this applies to as well as whether to show a confirmation before joining so you’re never caught off guard before joining your next call.

✨ New

AI Chat Text Settings: You can now control the text size and line spacing in AI Chat independently of the main window. See Settings → AI → AI Chat

🧪 New AI Experiments

Auto Model: Let AI choose the best model for the job at hand. We’ve enabled this by default so you can simply pick Auto when choosing a model and the most suitable model will be used for your prompt. We’re currently testing this (hence, experimental!) and will be fine tuning it as we receive your feedback. Please use the 👍 👎 and let us know how it’s going in #ai-experiments. Check out the AI Manual for more details.

Bring Your Own Models: For advanced users you can now add any OpenAI compatible LLM provider to Raycast AI. BYOM (Custom Providers) is disabled by default.

Note: To enable or disable AI Experiments go to Settings → AI → Experiments

🐞 Fixes

Root Search: Do not present AI Extensions popover for @ prefix
AI Chat: Fixed chat image capture layout
AI Chat: Ensure AI Chat window is not shown before screen capture
AI Chat: Ensure Escape cancels screen area capture
AI Chat: Starting a new chat no longer cancels current completion if chat branching is enabled
AI Chat: Fix image upload for custom models
Markdown: Fixed hang when processing HTML comments
Quick AI: Fix duplicate messages when regenerating a response
MCP: Improved compatibility with HTTP servers including Github and Vercel
MCP: Fixes issue running mcpm servers
BYOM: Do not pass temperature if not supported by the model
BYOM: Fix reasoning for Google API

🌿 Chat Branching (Experimental)

Create alternate conversation paths from any point in your chat history. Think of it as a "save point" where you can explore different directions without losing your original conversation. With background execution, you can kick off multiple branches that continue processing even when you switch between chats.

To use: From any AI Chat, simply press CMD + Shift + B to create a new branched chat and continue the conversation. Alternatively, you can right click on a chat in the sidebar, and select Branch Chat.

Note: This is an experimental feature. We’d like to see if you find it useful and how we might improve it. We’ve enabled it by default as it won’t disrupt your existing workflow, but if you’d like to disable it, go to AI Settings → Experiments. Please share your thoughts and feedback in #ai-experiments.

✨ New

• OpenRouter: We’ve added support for OpenRouter in order to give you even more control and flexibility over your AI usage in Raycast. OpenRouter provides access to a wide range of public models and has a beautiful, easy interface to manage your usage and costs without a subscription. To use: Add your OpenRouter API key in AI Settings → Custom Providers → OpenRouter. Once the key is validated, OpenRouter models will appear in the model picker.
• Clipboard History: Added a new Paste Sequentially command allows you to paste previously copied content one after another into your frontmost application. It’s best used with a hotkey, e.g. ⌥ + ⇧ + V.
• Calendar: Added support Riverside and StreamYard as conference providers for upcoming meetings.

💎 Improvements

• AI: Added a Show Tool Call Info switch in Settings → AI. When enabled you can view arguments and output for all tool calls in the chat
• AI: We now show a warning in the chat if there is an on-going incident with the LLM provider
• AI Chat: Composer text is now preserved when switching between chats
• AI Chat: Improved handling of presets in new chat menu
• MCP: The server version is now displayed when viewing a server in Manage Servers
• MCP: Added DENO_PATH and NODE_PATH to stdio environment

🐞 Fixes

• AI: Always use the original image when saving and sharing a chat image
• AI: Fixed cursor jumping after mentioning an AI Extension in root search
• AI: Fixed chat message layout corruption in Tahoe developer betas
• AI: Improved model fallback handling when using an API key
• AI Chat: Fixed jumping to first pinned chat after completion when >20 pinned items
• AI Chat: Generated chat titles should now use the user’s language
• Quick AI: Fixed Regenerate with Model action for commands
• MCP: Fixed the Streamable HTTP client accept header
• MCP: Fixed an issue parsing large numbers
• MCP: The HTTP client is now compatible with Atlassian’s MCP server

🔑 Bring Your Own Key

Bring Your Own Key (BYOK) seamlessly integrates your existing AI provider accounts with Raycast's intuitive interface. With BYOK you can now use Raycast AI with your own API keys for Anthropic, Google, and OpenAI. This allows you to send as many AI messages as you want at your own cost without a Pro subscription. To add an API key, open Settings -> AI and scroll down to the Custom API Keys section.

✨ New

• Manage Models: A new command to view and manage all your AI models. You can easily disable individual models or all models from the same provider using the command.
• Experimental AI features: The AI landscape moves quick — what works with one provider today might behave differently tomorrow, or might not be consistently supported across different models. You can now toggle on experimental AI features to try out early-stage features that we’re exploring, testing and trialing. Turn them on in AI Settings and join #ai-experiments in our Slack Community to share your thoughts!

Available Experiments:

• 🎏 MCP HTTP Servers: Support for HTTP MCP servers using the SSE (Server Sent Events) and Streamable protocols.
• 🦙 AI Extensions for Ollama models: If you’ve got Ollama installed, you can try out tool calling with local models. Since tool choice and streaming for tool calls aren’t supported by Ollama just yet, this can be a bit unreliable - which is why it’s experimental 😉

💎 Improvements

• Spotlight: Added new “Replace Spotlight with Raycast” command for a smoother transition to your favorite launcher
• AI Chat: Added new Send Active Window to AI Chat, Send Selected Area to AI Chat, and Send Focused Browser Tab to AI Chat commands to capture context even quicker

🐞 Fixes

• Safari Extension: Fixed an issue where the connection to the app would be closed and couldn’t reconnect. Make sure to update Safari Extension to the latest version in App Store.