18 September 2026
release 09-18-26Smarter default brains for every employee, two new premium brains, web search on every model, and clearer answers to your analytics questions.
Two more brains: Claude Fable 5.1 and GPT‑6 Astra New
Both are marked highest cost in the picker because they use more credits per reply, so they're there when the task deserves it, not as a default. Fable runs under an additional safety review by our cloud provider; details in our Terms, section 8.
Every brain can search the web Improved
You no longer have to choose Quick Search before asking about something current — the employee searches on its own when the question needs it, on every model. Quick Search is still there for research you want with proper citations: it uses Perplexity, which favours official sources and shows exactly where each fact came from.
Ask about your post performance, get a straight answer Improved
Ask any agent how your posts are doing and it now reads the numbers itself and answers directly — impressions, reach, engagement, laid out plainly — instead of dropping a raw data table into the chat.
- The AI Social Media Manager, the AI Receptionist and the Sagelyn Assistant start on Claude Sonnet 5. The AI Visibility Manager, which does the heaviest reasoning about your market, starts on Claude Opus 5. Employees you already have keep the brain they were created with; switch any time.
- Some accounts weren't receiving their monthly credit refresh. Credits now refresh correctly for everyone.
- Switching AI models in the middle of a conversation could lose everything said before the switch. Conversations now carry over correctly.
- Editing certain older agents could fail to save. Editing and saving now works for every agent.
- Uploading a file from your computer into the canvas could fail. Uploads now work correctly.
- If the invite email failed to send, the new teammate could be left with a half-set-up account and no way to log in. Invites now finish setting up the account first, so this can't happen.