AgentRQ growth: month over month and week over week
All 705 users since inception (2026-03-25), including the 5 heaviest. Data to 2026-10-22. October and the week of Oct 19 are partial, marked "to date". Updated to count only users who did real work, not anyone who triggered any event.
20
users did real work in October so far, and that number is flat. It was 18 in August and 17 in September. Users whose agent actually did something slipped from 16 to 14 to 13 (October to date). Signups also hold at about 50 a month, so growth is not coming from either side.
* October to date (22 days) compared with all of September, so changes understate the real trend. "Did real work" means a user created a task or message, or an agent made a tool call, completed a task or asked for a permission.
Only about 13 of October's 39 "active" users really used the platform
The earlier version of this report counted 39 active users in October: anyone with any recorded event other than creating an account. Here is what those 39 actually did.
The 19 users in the last group did not create a task, send a message or have an agent do anything. Their events were mostly "machine enrolment code created" (18 users) and "skill search" (10 users), which look like setup and browsing page loads. 33 of the 39 signed up in October, so setup-page visits by new users inflated the count.
Show data
Part 1 · Month over month
Signups spiked in June and settled at about 50 a month
New signups per month. June had 386, then 119, and since August it has run at 40 to 60.
At a straight-line pace, October would end near signups. October to date is not directly comparable with a full month.
Users doing real work are flat, and agent activity is slipping
Users per month who did real work, users whose agent did something, and users who completed at least one task.
Since August, users doing real work went 18, 17, 20 (to date). Users whose agent did something went 16, 14, 13 (to date). Users completing tasks rose from 5 in August to 9 and then 11, but the 5 heaviest users are part of every one of those counts.
Five users do almost all of the task volume
Tasks completed per month, split between the 5 heaviest users and everyone else.
Everyone, split by who did the work
Everyone elseTop 5 users
Everyone except the top 5
Tasks completed by the other 700 users: 7 in July, 8 in August, 42 in September, 47 in October to date.
The top 5 completed 92% of October's tasks (529 of 576) and nearly all of it before June. July is the exception: their volume dropped to 11 tasks, so the whole month looks empty.
Month over month, every metric
Each cell shows the value and the change from the month before. October compares a partial month with a full one, so its changes are understated.
Part 2 · Week over week
Weekly signups have held between 8 and 17 since August
New signups per week (weeks start Monday). The last bar is the partial week of Oct 19.
The last complete week (Oct 12) had 15 signups, up 25% on the week before (12). The week of Oct 19 has 8 so far, with about 4 days counted.
Weekly users doing real work have drifted down since late August
Users per week who did real work, users whose agent did something, and users who completed a task.
Users doing real work were 9, 8, 9, 9 in the weeks from Aug 17 to Sep 7, then 7, 5, 8, 7, 6 through the week of Oct 12. In the last complete week, 6 users did real work (down from 7), the agent did something for 5 (up from 4), and 3 completed a task (down from 4). The partial week of Oct 19 already shows 7 doing real work and 6 completing a task, so it may finish higher.
Tasks completed by everyone except the top 5, per week
The volume that comes from other users. It is spiky, but sits clearly higher since September.
Week over week, last 12 weeks
The last row is a partial week and is not compared as a full one. The full 31-week table is below.
Show every week since launch
What the trend says
Signups are not growing. They fell after the June spike and have held near 50 a month since August.
Real use is flat to slightly down. Users doing real work were 18, 17 and 20 (to date) from August to October. Users whose agent did something were 16, 14 and 13.
The earlier "active users are up 15%" claim was wrong. It counted setup-page visits by new signups. On the stricter measure, October is about level with September.
Depth is up modestly outside the top 5. Everyone else completed 47 tasks in October to date against 8 in August, which is a real gain from a very small base. First-time task completers rose from 3 to 7 to 8.
Volume is concentrated. Treat any total-volume number with care: five users account for about 92% of completed tasks.
Weekly numbers are noisy. A swing of 2 or 3 users reads as a double-digit change, so lean on the monthly view for direction and use weekly for spotting breaks.
Method and limits
All 705 users are included. The 5 users with the most completed tasks are split out only where marked.
Did real work: a user created a task or sent a message, or an agent made a tool call, asked for or received a permission decision, ran a scheduled task, or completed a task. Agent did something is the same list without the user-created task or message. Page views, searches, shortcuts, settings changes and machine setup do not count.
"Completed" counts tasks whose status is completed, dated by their last update. The schema has no completed-at column, so an edit after completion would move the date.
Event definitions come from the product's telemetry actions. A few rarely used actions may be misclassified, which shifts counts by one or two users at most.
Weeks start Monday and months are UTC. October is 22 days in and the week of Oct 19 is about 4 days in, so both are lower than a full period would be.
Counts only. No names or email addresses are included.