Changelog

What changed in Waper that you can see, newest first. Your agent reads this page too when you ask "where did this button come from?" or "why did this change?": if it isn't here, it tells you it can't see why instead of guessing.

November 2026

  • GPT-6.2 Sol: Pro, Max and the Sol you can pick directly now run on GPT-6.2 Sol at the same price; agents that had the previous Sol switch over automatically.
  • Your email and imported text: with Gmail or Outlook connected, your agent can search and read your recent email directly and links the original message; recently edited Google Drive documents, spreadsheets, presentations and PDFs, and what you import from Notion, Readwise and Granola, come into your library with their text. When you disconnect an app you can delete what was synced from it. See Apps.
  • Choosing a model: in your chat with an agent, a few specific models now sit below the three tiers (GPT-6.2 Sol and GPT-6.1 Terra, Gemini 3.8 Flash, DeepSeek V4.1 Flash, Claude Sonnet 5.5), each showing roughly how many times a Max turn it costs. Each agent keeps its own model, and Settings no longer has a default model; groups have no picker, and each member uses its own. Working out who answers in a group no longer uses credits. See Models and your own key.
  • It reaches out: in the daytime your main agent looks over your connected mail, calendar and what you're waiting on every so often, and only says something in its chat when it's worth it, with sources; the morning look is the first of the day. Tell it "stop messaging me first" or "less often" if you'd rather not; "not after 8 pm" changes its waking hours. It uses credits, shown as its own "Proactive check-ins" row in your usage. See When it reaches out.
  • Waiting on others: each agent's details have a "Waiting on others" section: what it keeps track of for you that isn't settled yet. Select Done or Never mind to clear one.
  • It sorts things out overnight: dates like a family birthday come from that person's page, whether or not you chatted that day; if you keep asking for the same thing by hand, the next morning it asks once whether to do it for you from now on; if you keep closing the same kind of suggestion, it stops suggesting it. Something a week past due gets checked against the original email before it is cleared, and when it can't tell, it asks you.
  • Dismiss a suggestion: a suggestion card on a message it sent you can be closed with ×, and it won't suggest the same kind again for two weeks; cards fold away at their deadline or after two weeks.
  • Routines: hourly routines can run only on weekdays or between certain hours; when a disconnected app blocks one it pauses, and it picks up again (and makes up the missed run) once you reconnect; a digest or report runs once on the spot when you set it up. See Routines.
  • Your instructions: every app and every Waper skill now has a "Your instructions" section on its page. Write how you want your agent to work (for example "put new Notion pages in the Work database") and it follows that every time. Say "from now on, always…" in a chat and your agent adds it there too. See Apps and Skills.
  • Your agent saves skills on its own: when you walk it through a way of doing something you'll do again, or correct it and say "always do it this way", it saves that as your skill and links it. Undo it if you don't want it.
  • Four new Waper skills: Scheduling, Inbox, Shopping and Travel. Each says what isn't its job.
  • Skills list the apps they use: a skill's page lists the apps it uses, and you can connect any that aren't connected right there. When your agent follows a skill and an app it needs isn't connected, it asks you to connect it first.
  • Tell the Waper team: when your agent can't do what you asked (an app or a capability is missing, or something keeps failing), it asks whether to tell the Waper team. Only if you say yes does it pass it on, in your words; any reply comes to your email.
  • The language of briefs and reports: when you ask your agent for a new watch or deep research and say "briefs in English", it writes in that language; otherwise it uses the content language in your settings.
  • Questions about Waper itself: ask about prices, plans or where a setting is, and your agent reads these docs before answering and links the page; ask about your own plan and credits, and it checks your account first.
  • One way to approve: there are no approval modes by the input box any more, just one way: most things are done directly, and it asks only before sending something to other people in your name, paying, or deleting something that can't be restored. See When it asks first.
  • Send cards: an email or a message going out in your name shows up as a card that is the draft itself; change the subject and the text right on the card and tap once to send, or just reply "send it" while the card waits. Tell it "no need to ask before sending expenses to finance" and it adds them to your trusted recipients on the spot.
  • Undo: it says in its answer what it did; to take something back, say "undo that" and it undoes it itself, edited documents included. Undo buttons stay only on lines where something of yours was deleted, moved or overwritten, where 5 or more things changed at once, and where who it may send to directly changed; "Remembered" is just a small line.
  • Talks more like a person: a simple question gets a sentence or two; for something that takes a while it first says what it is doing and then speaks up only when there is a result, and those lines stay in the chat; an answer may come as two or three bubbles. On Telegram each line goes out as it is said; by email it first replies saying what it is doing and replies again when it is done. See Progress and results.
  • Source chips: each site an answer used is a small chip with the site's name after the sentence, and a page already shown in the same answer isn't repeated.
  • Emoji reactions: react to an agent's message with an emoji, and it sometimes reacts to what you said; on the web, on iPhone and on Telegram.
  • Kiki speaks first: on your first visit, Kiki says hello first and tells you what you can hand it. A new agent you create gets straight to work when you give it a real task, and its duty follows from that.
  • Soul: every agent has a soul it writes in its own words and talks and works by; tell it how you'd like it to talk and it changes on the spot. Read, edit or roll it back under About in its details. See Soul.
  • People in your life, and memory: each family member, colleague or friend you mention gets a page in memory; when you correct or confirm how it works, it notes that in its working notes on the spot; ask "how do you know that?" and it finds the day and the chat where it was noted. With email or a calendar connected, at night it notes down what it read there about you and the people around you, each line marked with where it came from. See Memory.
  • Import memory from ChatGPT and others: in Memory, More → Import from ChatGPT and others gives you a message to send to ChatGPT, Claude and the like; paste the answer back and tick what to keep.
  • Hands-on on web pages, apps and iPhone: clicking, typing and filling in are done directly, and it stops for you at the step that sends, pays or deletes; when it's stuck on a step only you can do (scanning a sign-in code, say), the card says what to do, and tapping Done lets it carry on; when a page asks for a password, a code or a card number, a secure input card lets you fill it in without the agent seeing it. One iPhone does one thing at a time. See Your computers.
  • Chiming in, in groups: one member answers a message that names no one; when it touches something another member looks after, that member may add a line or two afterwards, or say nothing if it has nothing new. See Group chats.
On this page