July 27 - August 10, 2026

Feature highlights

• Routines that run on your schedule and skip what doesn't matter: routines can now repeat monthly or every N weeks alongside daily and weekly, and every trigger can be filtered on the details of the event itself, so "when a client meeting is recorded" no longer wakes up for your internal standup. Those filters and the new deterministic IF condition are decided instantly and for free, with no model call and no guesswork. Two new triggers join the list, when an email arrives and when an action item is created, and action nodes can now call a connected integration's real API action directly rather than asking an agent to do it.
• Routines fixes worth knowing: nested variables like {{input.meeting.title}} now resolve properly, a condition Revo couldn't read no longer quietly halts a routine halfway through while reporting success, and a bad time, an unknown timezone or a schedule with every weekday unticked is now rejected up front instead of producing a routine that looks active and never fires.
• The Brain graph, now a real part of Memory: Memory opens on a Graph tab built natively into the app, replacing the embedded view. Filter by entity type, isolated or ended relationships and date, all applied live, search by substring or ask a question and get an answer with citations, and open any entity or relationship for its aliases, provenance and supersession history. Sources are clickable links straight back to the meeting, document or issue they came from, and leaving Memory and returning brings the graph back instantly in the same layout.
• One plan per action item, written once: extraction, hand-off and planning used to each re-derive the same brief. Now a single planner writes one plain-English plan per item, and Claude, ChatGPT and Revo all work from that same artifact. If a fact is genuinely missing, Revo asks you a question in the item's chat instead of returning an empty plan, and the plan is shown to you as labelled, numbered steps rather than raw markup, while the editable version keeps its structure intact.
• Open in Revo now really runs it: picking Revo posts the plan into the item's own chat and the normal chat turn takes over, instead of playing a preview. ChatGPT joins Claude as a hand-off destination, and there's a "handle it myself" option that parks an item In progress with nothing running. Gemini has been removed, since it could never prefill a prompt.
• Hand-offs that agree with each other: every surface, board card, list row and focused card, now shows the same single primary action for an item. Your first "Open in Claude" or "Open in ChatGPT" quietly becomes your preferred assistant, and picking a different one later is treated as a one-off for that task, not a change of mind. Re-opening an item somewhere else updates the "Opened in Claude 5m ago" stamp instead of leaving it wrong forever.
• Handing work off no longer marks it done: an item sent to an assistant now rests In progress and stays in your active list until you mark it done or move it back to To do. You can snooze it, re-mention it without creating a duplicate, and Stop works during a live Revo turn, which is exactly when you want it. Stop also stops discarding your conversation, plan and artifacts once you've actually started working in the thread.
• Action item bugs, a long list of them: opening an item in Revo no longer sends the card back to To do or fails with "Cannot perform action", one failed execute no longer freezes skip, snooze, done and execute on every other card until a reload, dismissed cards no longer stay hidden after the action rolls back, the detail modal no longer silently swaps to a different item, creating an action item no longer shows an error toast when it worked, hand-typed items no longer spend a planning run on a brief you just wrote yourself, and simultaneous toasts stack instead of landing on top of each other.
• A board that behaves like a board: columns scroll independently, so scrolling To do leaves In progress and Done where they are. Dragging a card lights up the columns that will accept it, the placeholder matches the card's real height, headers carry a live count pill, and the card lands with a sweep. Every column has a Clear, and marking something done gets a small celebration across the board, list, focused view, home and the sidebar counts.
• Meeting recaps that lead with what to do: the recap email is now built from structured fields rather than one block of model-written HTML, so action items come first, then highlights, then the detail. Date, duration and attendees are read from the meeting itself instead of from prose, so they're right.
• One notetaker per call, and meetings that are actually yours: when two Revo teams are in the same meeting, a single bot joins named for both, and each team still gets its own recording. "My meetings" now uses the same attribution as the rest of the product, so a meeting you attended shows up there even when the calendar data is thin, and Upcoming meetings only lists events you're genuinely part of rather than everything sitting on a calendar you connected.
• An unpaid workspace stays readable: a lapsed subscription used to seal the whole workspace, including the invoices page. Now the paygate stops you adding rather than looking, so meetings, documents and issues stay browsable, billing, member removal and workspace deletion stay open, and only creating new work is blocked.
• Onboarding in the right order: you connect your accounts before choosing what Revo should do for you, so the capabilities screen can offer real per-mailbox options. A redundant settings step is gone, the capabilities screen matches its design with "Identify action items" shown as always-on, and the first-action screen shows your actual first action item as soon as there is one instead of the demo animation.
• A calmer home page: home is one centered column now, the Revo glyph, the composer, and a single tray tabbing between Actions and Suggestions. The composer doesn't grab focus on load, prompt suggestions appear only while you're typing, and action rows are down to title, time and the done/snooze/skip cluster.
• Integrations and documents: Microsoft Teams shows a native typing indicator while Revo works, matching Slack's eyes reaction. Slack rate limits and hiccups no longer break a run. Drive files show the right icon based on their real type, so PDFs and images stop looking like documents, visibility is editable in one click from the pill, and the crash on opening the Documents page is fixed. Sharing your Insights profile on LinkedIn now shows a proper preview naming your archetype.

How you will benefit

• Automate a real workflow: pick a cadence, filter the trigger, and let a routine call your tools directly.
• Explore what Revo knows about your workspace, and click straight through to where it learned it.
• Hand a task to Claude, ChatGPT or Revo from the same plan, and have Open in Revo actually do the work.
• Stop losing track of delegated work: it stays In progress until you say it's done.
• Read a meeting recap and know what to do from the first line.
• Keep access to your own data, and your invoices, if a subscription lapses.
• A board, home page and hand-off flow that behave the same way wherever you're standing.

Example

You set a routine to fire only when a client meeting is recorded, on a monthly review cadence, and file the summary into Jira through the connected action rather than an agent. After the call, the recap email lands with the action items at the top. You open the first one, read its plan as numbered steps, edit one line, and hit Open in Revo, which runs it in the item's chat while the card sits In progress. When Revo needs one missing detail it asks you in the thread. You mark it done, the board celebrates, and later you open Memory's Graph tab to see the client, the commitment and the meeting it came from, all linked.