- Faster search (3 PRs), p95 down about 40%
- New Slack threads support
- 12 bug fixes across billing
It already has the context
No re-briefing. Sauna pulls the thread and your notes, then drafts, because it remembers the decision you made.
ChatGPT starts every chat from zero. Sauna already knows your work. It remembers what you've done, connects to the tools you use, and does the task instead of describing it.

Every ChatGPT thread is a blank slate, so you keep pasting the same context back in. Sauna holds two kinds of memory instead: a durable record of your workspace, and a searchable history of what's been said. It picks up where you left off and acts on what it already knows.
Two memories that work together: what should stay true, and what was said.
No re-briefing. Sauna pulls the thread and your notes, then drafts, because it remembers the decision you made.
Sauna can research, run code, and hand back something finished. Ask for a board update and a Markdown doc comes back as a PDF, ready to send. ChatGPT tells you how. Sauna does it.
No more general LLM tools — it's just superior in every way.
Workspace notes hold what should stay true. Session memory holds what's been said. You stop re-briefing it every morning.
It reads files, runs code, and researches, then gives you something finished: a doc as a PDF, a draft ready to send.
Connect thousands of apps, with more than one account each if you need it. Anything missing, you add with an API key or an MCP server from chat.
No. Sauna remembers your work, connects to your real tools, and finishes tasks across Slack, email, and the web. It isn't a chat box you re-prime each session.
Anything that writes to your tools waits for your approval. Sauna drafts it; you decide whether it goes.
The web app, iMessage, Slack, email, and Superhuman. It's the same Sauna everywhere.
Sauna reads only what you connect, and acts only after you approve. Your workspace and its memory are yours, not training data.