Improvements

When creating a task with subtasks through the Sunsama connector, I want to supply each subtask's planned time in the same operation. On September 19, 2026, a ChatGPT connector test reported requesting a 17-minute parent and a 3-minute subtask. The returned subtask was 0 minutes; a later read through the official Sunsama connector in Codex still shows parent 17, subtask 0. The current create_task schema exposes only title and integration for each subtask. In contrast, add_subtasks_to_task accepts a subtask timeEstimate , and edit_task_time_estimate accepts a subtaskId . Please add timeEstimate to subtasks in create_task , or explicitly document the required second operation. If an unsupported estimate is supplied, please return a clear warning or error rather than leaving the caller believing it was applied. To inspect the gap: compare the subtask fields in those two creation tools, create a disposable parent with a subtask, and read back the subtask's planned time. Expected: an explicitly supported requested estimate persists, with the parent's total following the documented subtask aggregation rules. The original outgoing payload is not available in this review, so this is not yet a confirmed server-side write bug or a failure of edit_task_time_estimate . The 3-to-0 behavior is the reported creation-test symptom; the schema inconsistency is independently verified.

I'm finding myself recording an increasing amount of work running in the background with LLMs/Agents. I was thinking these tasks could sit at the top of my task list reminding me to occasionally check in on them and nudge them forward, rather than be planted in a specific spot on my calendar. Bonus points: when my agent has been blocked by me for >X minutes it can ping Sunsama to move it to the top of my list.

Would love to see an integration with something like Granola to automatically capture notes and tasks from a call. This would be incredibly helpful and allow me to ensure every action item from a meeting always makes it to Sunsama.

I click the icon in the bottom right, I click "Live", and I immediately start describing a task that I want created. After I get done speaking, Sunny says something to the effect of "Hi, I'm Sunny, let me help you plan your day". I thought it had completely ignored what I had said. I said "really? I just told you" and it responded "Sorry, I see you wanted to create a task to <XYZ>. I've added it to your backlog". It did what I asked it to do, but the standard greeting happening after I started the conversation was confusing and frustrated me at first.

Hi! As a college student with ADHD, I often find manually inputting my syllabi into a planner so tedious that I just end up not doing it at all. I'd love a tool that makes this automatic. Here's sort of what I'm thinking: Upload a syllabus (or multiple syllabi) Choose a default lead time for reminders (e.g., "2 days before due date") Have the tool automatically parse each assignment and create it as a task/subtask on my calendar or kanban board, tagged with: Class name as a tag Actual due date in the date, but added two days before in the kanban Example: I upload my syllabus and select "2 days before due date." The tool then creates a task (and subtasks) for each assignment labeled with the class name, showing the real due date, but scheduled to prompt me 2 days early. Thank you so much!

The ability to capture tasks by voice directly into Sunsama on mobile in a frictionless way is needed. Though, as I write this I realize that capturing tasks via mobile doesn't even have a dedicated UI, i.e. the "companion" app just uses the iOS UI if you place a widget on the homescreen, etc. so the chances of getting any improvements to the mobile experience is as rare a prospect now as it's ever been.

work email has write access to shared calendar, and it accidentally deleted a shared event. This should be prevented with some guardrails

Ideas for tasks I should do often occur to me while I am running. My current workflow is tap to open Sunsama. Tap add. Tap in the text entry field. Tap Wispr Flow to record. Tap to stop recording. Then tap to accept the text that Wispr Flow entered. I like using Wispr Flow in these moments because I'm often out of breath and have trouble putting my thoughts together on a first pass. It would be great if Sunsama had a native feature for 1-touch voice entry that included some basic LLM processing to turn messy inputs into more clear / concise entries (a la Wispr Flow). The ability to enter a task via voice without all the fiddliness and button-tapping of my current workflow would be a huge win!

The same AI integration that exists on website/desktop could exist on phone where it's useful and accessible in daily life.