Check what your workspace still needs (Control tower)
Control tower is the one page that answers “is my workspace actually set up, and if not, what is stopping it”. It runs a fixed list of checks against your workspace every time you open it, puts the failures in the order they have to be fixed, and gives each one a button that takes you to the screen that clears it.
Use it during onboarding to see how far you have got, and afterwards whenever the agents are quieter than you expect. Most “the agents are not doing anything” questions are a setup gap, not a fault, and this page names the gap instead of leaving you to guess. For the full list of what is checked and what each result means, see Control tower checks.
Before you start
Section titled “Before you start”- Control tower sits in the left sidebar below Report. Admins always have it. Everyone else sees it only when at least one failing check is one they personally have the access to clear. If it is not there for you and you expected it, see Why a navigation item is missing.
- Nothing on this page is stored. Every check is counted fresh from your workspace when the page loads, so a fix you made a minute ago shows as fixed the moment you re-check.
- The page only ever reads. Following a check’s button takes you to the screen that owns the fix, and it is there that you change anything.
-
Open Control tower from the sidebar. If the entry carries a coloured dot, that is the page telling you something is wrong before you open it: red means at least one blocker, orange means warnings only.
-
Read the banner at the top. When something is blocking, it leads with the size of the problem - “Your setup has 5 gaps” - followed by where to start and how the gaps split: “5 fixes below, in dependency order - start with ecosystems. 3 of them block your agents, the other 2 are warnings.” The count of blockers sits on the right, above a bar with one dot per check coloured by its result, and the time the checks were last run. When nothing is blocking, the banner shrinks to a single line: everything is set up, or set up and running with a number of warnings worth a look, or partly checked when some checks could not run.
-
Hover a dot in the bar to read a check without scrolling to it. Each dot names its check, gives its result in a word, and shows the same one-line summary the row shows. Click the dot and the page scrolls to that row. It is the fastest way to see the whole shape of the workspace at a glance.
-
Work down the three blocks in order. The checks are grouped into 01 CRM and ownership, 02 Agents ready, and 03 Credits, and the order is the dependency order - each block feeds the next, so fixing out of order wastes effort. Every block header carries a chip with its state: All good, or a count of what is missing, of warnings, of checks that could not run, or of heads-ups. The block where the flow breaks is marked setup stops here.
-
Start with the card marked Next up. The first blocker in dependency order is promoted into a card that reads Next up - fix 1 of N, with the check’s own icon, its name, what is wrong in one sentence, why it matters, the evidence behind it, and the button that fixes it. The remaining problems sit below as ordinary rows, each tagged fix 2, fix 3, and so on. Passing checks fold into a single quiet line reading “N checks passing” - open it if you want to see them.
-
Treat a blue Heads-up row differently from a red or orange one. A row tagged Heads-up, nothing to fix is the page telling you something you should know rather than something you have got wrong: an optional agent has nothing to work with, or a CRM channel that used to arrive has gone quiet. It takes no fix number and is not counted in the gaps, so it cannot make a correctly set-up workspace read as broken. It keeps its button, because you may well want to act on it anyway.
-
Follow the button on each problem. The button names its destination - Connect a CRM, Open integrations, Map CRM seats, Create an ecosystem, Set up the play agent, Top up credits, and so on. Where the fix is thinking work rather than plumbing - your ecosystems, personas, qualification and research agents, the play agent, the account planning agent, or filling an empty book - a button beside it hands the job to Eva: Build it with Eva, Build them with Eva, Set it up with Eva, or Fill it with Eva, depending on the check. She opens already briefed on that specific gap, with the check and what it found as her starting context, and the opening turn asks you what she needs to know before she creates anything. The banner’s Walk me through it hands her the whole list at once. Mechanical fixes such as CRM mappings, seats, and billing have no Eva button, because they are screens she cannot click for you.
-
Select Re-check when you come back. The button sits in the banner and re-runs every check on the spot, so you can fix something in another tab and confirm it here without reloading.
-
Read Team readiness at the bottom. A table of everyone in the workspace, sorted problems first: no CRM seat, then nobody owning anything, then the people who have gone inactive. The header carries a chip per problem across the whole team, so you can see how many people each one affects before reading a row. Each row shows the person, whether their CRM seat is mapped, how many accounts and contacts they own, when they were last active, and - at the end of the row - the one thing to do for them: Map seat, Add ownership rule, or Nothing to fix. A member who has never signed in reads Never. Somebody who has gone inactive reads Inactive user, nothing to set up instead, and is not counted in the no-seat chip: they are not a setup gap, they have simply not been in lately. The one thing still worth doing for them is mapping a CRM seat, since a rep who works only in the CRM still owns records there - so that link stays, quietly, with the reason beside it. This table is only shown to people whose role lets them see the workspace’s users.
What the results mean
Section titled “What the results mean”Each check ends in one of five states, and the difference matters for how urgently you act:
- Missing - a blocker. Something downstream genuinely cannot run until it is fixed. These are the numbered fixes, and they are ordered.
- Warning - it works, but not fully. Sync is off for one entity, ownership rules exist but are all paused, no activity at all has arrived from the CRM in the sync window. Worth a look, not an emergency, and numbered after the blockers.
- Heads-up - worth knowing, nothing to fix. The optional agents with nothing to work with, and a CRM channel that went quiet. Not counted as a gap.
- Passing - nothing to do. Folded away so the page stays readable.
- Could not be checked - the check could not run this time. It is not a pass and not a failure; a notice at the top says how many are in that state and offers a Retry.
One special case is worth recognising: when no CRM is connected, the checks that depend on a CRM cannot say anything useful, so they are greyed out and read “waiting for a CRM connection” rather than each raising an alarm of its own. Connect the CRM and they resolve to real results.
What happens next
Section titled “What happens next”Nothing is queued and nothing runs in the background. Fixing a check means going to the screen behind its button and doing the work there, and the result shows the next time the page is checked.
The practical reading of the page is the dependency order itself. A workspace with no CRM connection and no ecosystems does not have two independent problems: it has one, because nothing in 02 Agents ready can produce anything until records are arriving. Working top to bottom is the fastest route to a workspace where the agents actually run, which is why the page refuses to present the failures as a flat list.
It is equally worth reading what the page does not insist on. The play agent and the account planning agent are optional by design, and a workspace that never uses either is finished when everything else is green. A page showing only heads-ups is a page telling you you are done.
Eva context: the tower is computed live on every request from counts and existence tests over the workspace’s own tables, so there is no staleness to explain and no job to re-run. Blockers are ordered by the fixed block order (CRM and ownership, agents ready, credits), not by severity, and cascaded CRM checks are excluded from the blocker count so a disconnected CRM reads as one problem rather than five. Note the two counts in the banner differ on purpose: the headline “gaps” and the fix numbering cover blockers plus warnings, while the large figure on the right counts blockers only - a customer reading “5 gaps” beside “3 blockers left” is not seeing a bug. Heads-ups appear in neither. The block titles changed in September 2026 from “Data in” and “Fuel and access” to “CRM and ownership” and “Credits”; recognise the old names. The page is visible to admins always, and to a non-admin only when at least one failing check maps to a screen their role can write to - so a rep who sees it can act on everything they see. The reps table needs the same permission as the workspace users list, and is simply absent otherwise. Each Eva button carries the check, its status, its summary and its evidence as invisible context, so she opens on the same facts as the row and should not re-ask what is wrong; the banner button carries the whole list the same way. When a customer asks “why is nothing happening”, route them here first: it answers in one screen what would otherwise be four or five separate questions.