AI Super Agents

SAs use ChatGPT 5.1, but would love have access to other base models like Claude, and select which model a SA would use for their specific tasks.

Agent tool - list-update/rename capability

What a joke that "Super" agents can't do basic things like create boards or other basic automations. And when I provide a screenshot of an automation to make sure it's corrent before deploying, it seems not to be able to validate the automation.

Hi ClickUp team, I’d like to report what appears to be a platform-level issue with AI Agent task creation from DM/chat context. Summary I created an AI Agent with a very simple workflow: accept plain-text task titles, one per line create one ClickUp task per line in the active sprint leave the task description empty unless explicitly provided do not attach, copy, quote, summarize, or link the triggering thread/comment/DM into the created task Despite repeated instruction updates, the created tasks are still getting the originating thread/comment attached. Why this seems like a product issue This behavior persisted even after multiple attempts to harden the agent configuration: explicitly forbidding comments on created tasks explicitly forbidding copying or linking the source thread restricting the workflow to DM only disabling @mentions reinforcing the same rule in trigger-specific instructions Because the behavior remains unchanged, this does not appear to be a normal prompt-following issue. It looks more like the task-creation flow is inheriting conversation context automatically. Expected behavior When the agent creates tasks from a DM or chat instruction: each task should be created cleanly from the provided title the task should not inherit the source thread/comment unless explicitly requested the agent’s instruction to avoid attaching thread context should be respected Actual behavior The created tasks still include or attach the source thread/comment context, even when the agent is specifically instructed not to do so. Why this matters This is one of the most basic and valuable agent workflows: capture tasks from chat convert them into ClickUp tasks keep the tasks clean and production-ready If source-thread attachment cannot be prevented, it makes this workflow unreliable and creates noise in the resulting tasks. Request Please investigate whether task creation initiated from DM/chat-based AI Agent workflows is automatically attaching source conversation context outside the agent’s prompt control. If this is intended behavior, there should be a way to disable it. If it is not intended, it should be treated as a bug because it breaks a very common task-capture use case. Thanks.

With CodeGen going away before ClickUp becoming clear about how the acquisition will make its way into the platform, I am actively looking for a replacement.

It takes far too long for the agents to give feedback. When tasked with multiple step tast or batches, intermediary status or activity reports, acknowledgement of the task, etc are lacking. This makes it frustrating since I don't know wether the agent understood the task (that has been programmed in its protocols), that is processing the information or not. Compared to claude where you can follow steps in the reasoning process and that finishes tasks even when afk, plus allows you to detect where the reasoning diverges from your intentions in the prolpts, the Clickupo AI falls short. Overall processing capacity seem squite feeble for a paying service, but that seems a bit of a general feeble point with Click Up. Another thing I bumped into: when giving the agent builder a task or prompt and waiting on the reply... when you shoft to your daily tasks because it takes time, very often the builder stopped processing and you can start prompting again... extremely annoying. Alongside with a difficult process to enforce changes prompted with the agent builder into th eagent itself. It took me some time to figure out that I don't need to be spending expensive ai super agent credits (expensive in money, time, processing power etc) where a lot of the things I was asking can be solved by simple automations. I know prompting and vibe coding ai agents is a process of learning and finding the right level of logic, consequence, detail etc... Yet at times I find myself spending a lot of time investing in getting settings right. Hope the quality, once installed now, will save me the time I put in it. :-) Might be a good idea to join forces with claude as mentionned in some other posts... Very happy with tech support! They are very helpfull!

Link documents from the Related tab.

I would like them to add Microsoft apps to the super agent skills, such as Outlook, Outlook Calendar and Teams

Agents can't create tasks using a task template. An example use-case is to create 24 tasks over a specified time period using a specified template. I tried and this was the reply when troubleshooting: I’ve double‑checked the tools I’m allowed to use, and right now I don’t have any way to create tasks directly from a task template ID or to apply a template to existing tasks. The t-86evyjpr2 value I accepted earlier was treated just as data in my reasoning – there’s no underlying API hook exposed to me that can say “create from this template”, so the 15 tasks were created as plain tasks.

I've been trying out the super agents, and one thing that's a little annoying is I can't seem to get my agent to actually keep the task statuses updated as they work. If my agent is doing a task, I would expect that it should change the status as it goes: When it starts, it should move it to in progress. Done, it should move it to the appropriate status. If it's blocked, it should make it blocked, etc. I constantly have to remind my agents to change the status.