v1.3.0 — Smarter models, programmable workflows
Our biggest release since launch just landed. The highlights:
Two new flagship models
- Claude Opus 4.8 is now the default Claude tier — our most capable model yet, with a larger context window so it keeps more of your documents and conversation in mind at once
- Fable 5 joins the roster for users on paid plans
Workflows that actually compute
Workflows used to run a fixed, linear set of steps. Now they branch, loop, and remember:
- Iteration — run a step, or a whole branch, once per item in a list
- Switch & limit — route down different branches based on a value, and cap how many items flow through
- Cross-run data store — key/value memory that persists between runs
- Expressions — reference and transform earlier results inline, anywhere in a node
- Declared run inputs — a workflow asks for what it needs through a clean guided form
- Context agent — a side-panel helper that builds and debugs alongside you on the canvas
A real reading room
- Conversation artifacts → Library in one action
- Reading mode renders long documents in a comfortable, article-style layout
- Open in a new tab for any Library item
- Community posts now support inline images and render as clean, readable articles
Google, surfaced
When an agent works with Gmail, Drive, or Maps, results now come back as tappable link cards — and location answers render as interactive map cards grounded in real map data.
Everyday polish
Sign in at /login and sign up at /signup with shareable links, a redesigned composer toolbar, a new Career & Job-Search specialist team, and full chat history in the list.
Full notes
The complete release note lives in the in-app What's New drawer. Open it from your avatar menu in the top-right.
A lot of this came from you
Several items in this release came directly from posts in Feature Requests and Tips & Tricks. Keep them coming — we read every one, and we tag the ones that ship with a link back to the original post so you can see the loop close.