What the agent does
This is for a video production studio that makes brand films, product videos and social cutdowns for a dozen clients at a time. Every deliverable is a ClickUp task, and when an editor sends a cut, the task moves to a Client review status and waits. Some clients reply within a day; others go quiet for two weeks and then ask why the launch video is late. The producer wants a list every Monday of everything waiting on a client, oldest first, with a nudge ready for anything past five working days.
The studio connects ClickUp once with POST /me/connections {"type": "mcp", "name": "clickup", "url": "https://mcp.clickup.com/mcp"}; ClickUp's docs say OAuth is the only way to sign in and the server is on every plan. A run finds the Client Work space and its client folders with clickup_get_workspace_hierarchy, then calls clickup_filter_tasks for tasks in the Client review status across the space, which ClickUp's docs say filters by status, list, folder, assignee, tag, date and custom field and returns results in pages. For how long each task has sat in that status it calls clickup_get_bulk_tasks_time_in_status with up to 100 task IDs at a time, which needs the Total time in Status ClickApp turned on.
For each task it reads the last few comments with clickup_get_task_comments to find the last thing anyone said: the client's note, the editor's message with the review link, or nothing. It groups tasks by client folder, sorts each group by days waiting, and for any task past five working days drafts a short nudge from the producer naming the deliverable, the date the cut went out, the due date it holds up when the task has one, and one line asking for notes or an approval.
It changes nothing in ClickUp. ClickUp's docs mark clickup_update_task and clickup_execute_operator as writing, and clickup_create_comment and clickup_send_chat_message as posting as the signed-in user; the prompt forbids every tool that writes. Clients see nothing until the producer sends a nudge from their own email.
Start it with one API call
POST https://api.phoenix.so/o/{org}/sessions with this body:
{
"agent": "claude",
"input": "List every deliverable in our ClickUp space Client Work that is waiting on client review, as of 2026-11-16. Use the clickup connection. Call clickup_get_workspace_hierarchy to find the space and its client folders, then clickup_filter_tasks for status Client review across the space, every page. Call clickup_get_bulk_tasks_time_in_status for those task IDs, at most 100 per call, for each task's time in Client review; if that is unavailable, use the date of the comment that shared the cut and say which you used. For each task read the last 3 comments with clickup_get_task_comments. Group by client folder, oldest first: task name, assignee, days in Client review, due date if set, last comment (who, when, one line). For each task over 5 working days in Client review, draft a nudge under 70 words from the producer: name the deliverable, when the cut went out and the due date it holds up, and ask for notes or approval. Write review-queue.md. Do not call any tool that creates, updates, moves, comments, sends a message or runs an operator.",
"connections": [
"clickup"
]
}- Get an API key. In the console, open Settings, then API keys, and create a key for your org. Send it in the Authorization header as a Bearer token.
- Send the request. POST the body in request.json to https://api.phoenix.so/o/{org}/sessions, with your repository as the resources URL. The run starts in its own cloud sandbox.
- Read what came back. Follow the run on its live stream or wait for the run.completed webhook, then read the result and its trajectory.
What comes back
The session returns review-queue.md as an artifact: a section per client, each waiting deliverable with its days in review and last comment, and a drafted nudge under each one past five working days. The answer names the deliverable that has waited longest.
What the producer checks: tasks left in Client review after the client approved by email, which show as waiting until someone moves them; nudges for clients the account lead is already talking to; and the days-in-review figure, which comes from ClickUp's time in status only when the ClickApp is on, and from comment dates otherwise, as the list says for each task.
Run it on every Monday morning
The producer saves the agent with POST /o/{org}/agents, naming the clickup connection, with the space, the status name and the five-day rule as its instructions, and adds a console automation with a schedule trigger (0 9 * * 1, timezone set to the studio's) that runs it. A schedule runs the saved prompt as written, so the saved prompt says as of today in place of a date.
In a busy launch month the producer can add a second schedule trigger on Thursday. Each run is its own session with its own list, so Monday's stays as it was.
The use case this job follows, with its automation and trigger: Produce a report or file, no repo.
Why run it on Rush?
ClickUp signs in only through OAuth, so the sign-in lives in Rush's connection rather than in a script's settings, and every task and comment the agent read is in the session's transcript. When a client says they never got the cut, the producer can find the comment that carried the link and its date.
ClickUp's docs limit MCP calls per workspace per rolling day on plans without its AI add-on, from 100 on Free Forever to 1,000 on Business. The run reads time in status for up to 100 tasks in one call and reads comments only for tasks in review, and each session records its calls and model cost, so the studio can see what the Monday list spends of that allowance.
Frequently asked questions
How many ClickUp calls does this use?
Each tool call counts as one against ClickUp's daily MCP limit; ClickUp's docs say initializing and listing tools do not count. The list costs a hierarchy call, the filter pages, one time-in-status call per 100 tasks and one comment read per waiting task.
Can the agent post the nudge as a ClickUp comment?
It could with clickup_create_comment, which notifies anyone mentioned, but this job keeps that with the producer. Many clients are not in the studio's ClickUp at all, and a nudge reads better from a person's own email.
What if we do not use the Total time in Status ClickApp?
ClickUp's docs say the time-in-status tools need it. Without it, the prompt has the agent use the date of the comment that shared the cut, and the list says which figure it used for each task.
Does it see every client's work?
It sees what the person who connected ClickUp can see. The prompt names the Client Work space, so internal spaces stay out of the list even when that person can open them.