MCP tools: Work log
Last updated 18 days ago
MCP tools: Work log
Liza keeps a work log: who worked on what and when — tasks, lists, chats, threads, calls and calendar events, grouped into work sessions with a result and a next step. The tools do not return raw data but a pre-filtered selection: for "what did I do yesterday?", "which task did I mean?", stand-ups and picking work back up.
Visibility
You always see your own entries, including private ones (only opened, unpublished AI runs). Other people's entries only if they are shared, contain a contribution and belong to a task you can see yourself — or to a chat, thread, call or event you take part in. If the organization has switched the work log off, the tools answer enabled: false instead of an empty list.
People (user) are given as "me" (default), a name or a user ID; an ambiguous name returns the candidates to choose from. Times of day (daypart) are in the organization's time zone: early before 10:00, morning 8–12, noon 11–14, afternoon 13–18, evening after 17:00 — the overlap is intended.
recall_work
Find the object someone means from a vague memory — "the task I created this morning after talking to Anna". read · follower
Response: ranked candidates with a reason ("created 08:42, 6 min after the call with Anna"). If clear_match is true, have the top candidate confirmed; otherwise name the two best.
get_work_timeline
What did someone work on during a day? The work sessions in chronological order, with object, start, result and next step. read · follower
get_object_history
Who did what on this object? The sessions, newest first, with person, time, summary, result, next step and milestones (e.g. "blocked: waiting for API access"). read · follower
resume sums up re-entry: who worked on it last, where to continue (the latest next step) and whether it is currently blocked — meant for taking over a task or coming back after a break.
get_work_digest
Prepare a stand-up, a day review or a weekly review from the work log.
read · followerThe response comes as lists: done, progressed (with next step), blocked (with a note), missed/cancelled and open blockers with their age; each object appears once (sessions = number of sessions). For a stand-up it also carries planned_today: at most 8 tasks with a reason (must_today, planned, due, in_progress). If more_in_progress is set, more tasks are in progress — then propose 3–5 priorities and put them on the plan with plan_for_day. The AI writes the stand-up text itself from this data.
plan_for_day
Put a task on the personal day plan — it then shows up in the cockpit under "Today" or "Upcoming". write · follower
The task itself does not change — no due date, no status.
Set must only after consent
The must-do mark triggers reminders. Only set it once the user has agreed, e.g. after you proposed priorities.
Was this helpful?
Still need help? Ask the team