Checking your sign-in…
STRIDE Admin help Admin console User help
www.strideos.me

Administering the system

For whoever runs this installation: adding and removing people, organizing teams and boards, and the settings that decide who gets in.

The console is not the security boundary

The Admin button only appears for a God Mode holder or a System Admin, but every action is checked again by the database when it runs. A non-administrator who reaches this page is refused on every operation.

Admin consoleadmin.html

You reach it from your account menu, top right; the entry only appears for a God Mode holder or a System Admin. The console wears the same top bar as Tasks, Notes, Team and OKRs — search, Refresh, Messages, theme and your account button all work the same way. Create, Today and Focus are missing on purpose: none is an administrative action, and the one thing this console creates — a team — has its button in the section that owns it.

Two roles open it, and they see different consoles. A God Mode holder sees every section below, for every organization. A System Admin administers one organization and sees only Overview (that organization's figures), Organizations (its own card, with no Create organization), Teams, Boards, OKRs, People and Audit, each limited to that organization, with its name in the header. Everything else on this page is for God Mode only. See System Admins and God Mode.

Getting around: the section list

Down the left: Overview, Organizations, Teams, Boards, OKRs, People, System and Security, Email configuration, Live editing, Human verification, API bench, Storage Explorer, User traffic and audit trail, Audit, Browser errors, Feature usage, and Distributed builders. The list is grouped under four headings: Overview, Directory, Configuration, and Data and records. There is no registration section — anybody may create an account on the sign-in page, and what proves it is the six-digit code emailed to that address, which nobody here administers; see Creating an account. A faint sentence under the heading always says what the current area governs and that it applies to the whole application, not one team — the same words (teams, boards, people, email) also name settings that belong to a single team, a few clicks away on the team board. The section you were last in is remembered on this computer. On a narrow window the list becomes a scrollable row of buttons above the content.

STRIDE My WorkTasksNotes CalendarTeamOKRs Messages Search tasks, notes, people, teams, OKRs…  ⌘K ☾☀ JD
Sections
Overview Teams 2 Boards 5 People 12 System and Security Email configuration Human verification Storage Explorer Audit 18
Overview — the whole application. One team’s own settings live on the team board, not here.
2
active teams
5
boards
12
people
—
files awaiting removal
Configuration at a glance
Application email · on
Password strength · High
Human verification · on — registration, password reset, repeated failed sign-ins
Malware scanning · not performed — there is no switch

Ask STRIDE on the console

The console has one Ask STRIDE button, in the top bar, and ⌘/ (Ctrl /) opens the same panel. The panel is set to the section you have open: a chip such as Admin Console › People says so, it follows you when you open another section, and × on it clears it. The starter questions are written for that section. The section only decides which how-to notes are tried first; you can still ask about anything, and the answers are the ones written for system administrators. Show me and Take me there go to the section's button in the list on the left and ring it. Only the name of the section is sent, never an account, team, address or setting shown in it. What else is sent, and what is kept, is in Ask STRIDE, and what it sends.

More than one administrator at a time

The console does not re-read itself in the background — press Refresh after someone else has been working, or before acting on a list you have had open a while. This is deliberate: the console's lists carry ticks, and an automatic repaint would clear a selection mid-choice. If two of you are in here at once, refresh before you act — an account someone else just deactivated still shows active until you do.

Overview: the application at a glance

Where the console opens. It answers three questions and changes nothing — every figure is a door to the area that owns it.

There is no security score, no percentage of your quota, and no per-team health grade — none of the three is something this system measures.

Organizations

Every organization, with its team count, people count and how many Organization Owners it has. It is not read-only: its row's Manage owners button opens the roster, offering Make owner / Remove as owner on each person, and the section offers deleting the organization. Neither needs System Admin — the database checks Organization Owner or God Mode, exactly as it would from Organization settings, so a System Admin who is neither sees the same buttons as anyone else and is refused by the database on pressing them. A System Admin sees only its own organization here.

In the figures at the top, Teams, Boards and People are doorways: press one to open that section. The others only report a count.

Every count matches its list. The browser tests make their own organizations, teams, boards and accounts and delete them when they finish. The console leaves out the ones you are not on, in its lists and in every count: the section list’s numbers, these figures, the Overview, and the People, Teams, Boards, OKRs and Live editing sections. People here is the number of accounts the People section lists. To include test data, turn on Show test boards: at the foot of the organization switcher, or in My Settings → Workspace → Test workspaces. User traffic and the audit trail still record what test accounts did.

Create organization (God Mode only) asks for the organization's name and the name of its first team, because every organization needs at least one team. Leave the team empty and it says “Enter a first team name to create the organization.”

Deleting an organization refuses while it still holds a team or anyone but you — delete its teams first, or remove its people from Organization settings. Once empty, Delete asks you to type its name.

An organization that still holds teams and people can also be deleted all at once — including accounts

When the plain delete refuses, the dialog offers Delete everything in it — its teams, boards, tasks, OKRs and every membership — previewed and counted first, and permanent. Keep the people's accounts is ticked by default: ticked, everyone only loses membership of this organization; unticked, every account in the organization is deleted outright with everything it owns, including work unrelated to this organization. The preview names how many accounts that is.

This path needs God Mode: an Organization Owner alone can only delete an organization that is already empty.

There is deliberately no Inactivate for an organization — unlike a team or board it has no reversible closed state. An organization is otherwise configured — name, address, mark, billing, its own invitations — from Organization settings, by an Organization Owner or God Mode holder; this console only adds ownership and deletion.

An empty list, or a section saying the Organization layer is not yet deployed to this database, is an honest answer: no organization rows exist yet.

Live editing: the system switch and the organization switches

Organization settings holds a Working together card with one switch, Live editing. When it is on, people in that organization see each other’s edits to shared notes, task checklists and comments as they type. Only the organization owner can change it from there, and the database refuses anyone else. Switching it on or off reaches notes and tasks that are already open within about a minute, and everyone keeps saving normally.

Live editing is also a section of this console, directly below Email configuration. It has two parts, and you need to be a STRIDE administrator to use them.

A row says Live when people there are editing together, Off when its switch is off, and Held when its switch is on but the system switch is off. You can set organizations up while the system is off; they go live when the system comes back on.

With live editing off, everything saves the way it did before live editing existed. Nobody is reloaded: a colleague’s changes appear when the note or task is reopened or refreshed, and a save that would overwrite a newer change is refused instead of being lost.

The organization owner can change their own organization’s setting too, and both places change the same setting, so they always agree: the last change wins, and every change is recorded in Audit. The owner’s card says who last changed it, naming “a STRIDE administrator” for yours, and reads Off for STRIDE while the system switch is off. Readers see a small Live editing ready (gray, alone), Live · 2 here (blue, with someone else) or Live editing off label on shared notes and tasks; when it is off because of the system switch, an administrator’s label offers Open Admin Console → Live editing. On Overview the Live editing line reads On · N of M organizations while the system switch is on, Off while it is off, and a dash until its figures have loaded.

People

The section is called People, because that is what is in it. It was called Users, and a team's line read "4 members"; both were the database’s words for a human being. Every registered account, with how many tasks, teams and boards each has.

God Mode — visible only if you already have it

God Mode is a separate, product-wide flag — not a team role, and not System Admin. A God Mode holder can see every organization, team and board in the product. It is configured only from this People section, and only an existing God Mode user can grant or remove it. If you do not hold it, the column and control are not shown to you at all. God Mode alone opens this whole console and everything in it, so a holder needs no other role, and any holder can grant it to another account.

A row carries a God Mode badge when the person has it. Opening their Profile shows the full control, with a confirmation before anything changes. You cannot change your own status, and the product refuses to remove the last remaining God Mode user. A God Mode reader sees (God Mode) appended to the browser tab title on every page, so it is never silent.

It also runs invisible the other way, in every organization the holder is not a member of: there a God Mode holder is never listed in that organization's own People, counted in its user totals, or shown on any of its rosters, and it holds no membership to leave or transfer, so no Leave-Organization control shows for it. God Mode adds reach, never membership. A holder who has also joined an organization the ordinary way (on one of its teams, or as a member or Owner) is listed there for that membership like anyone else, and can leave or transfer it like anyone else; God Mode is what lets one holder belong to more than one organization at once. This Admin Console is the only place any God Mode grant is ever visible, to another administrator — a God Mode holder; a System Admin never sees one. Every grant and revoke is written to the same record System Admin changes are, and Audit reads it back to a God Mode holder, labeled as a change between user and god_mode.

Granting or removing God Mode asks for an emailed code. After you confirm, STRIDE emails a 6-digit code to your sign-in address and shows a Code box beside the control, with Confirm, Send a new code and Cancel. A code confirmed in the last five minutes is used again without asking; Cancel changes nothing. Every God Mode holder, and the person changed, is told.

Builders access — given by God Mode to one System Admin

On a System Admin's Profile, a God Mode holder sees a Builders access section with Give Builders access or Remove Builders access. It asks for the same emailed code. A person with Builders access opens Distributed builders read-only, and their prompts there are requests that a God Mode holder approves before anything ships. Nobody changes their own, a person with no System Admin role cannot be given it, and it ends when they stop being a System Admin or the account is deactivated. Only God Mode holders see this section.

Narrowing the People list to one team

Three controls narrow the list together: Team (starts on All teams), Is System Admin, and Search people… (name or email). Clear filters resets them and only appears while something is narrowing the list. Someone in two teams appears under either, once; someone in no team appears only under All teams. All three compose with the active/inactive filter.

The figures above the list filter it too: All accounts, System admins, Dormant, Inactive (no sign-in) and (when the Organization filter is showing) No organization. Each sets the same control as its switch or dropdown, so the two always agree; System admins and Dormant can both be lit at once. Press a lit figure again to clear it, or All accounts to clear everything. Pressing one clears your selection, and Select all takes only the accounts shown. Dormant is dimmed, with the reason, when dormant accounts cannot be read.

The filter narrows what you can already see

It is a convenience, not a permission — it can never reveal an account the filter was hiding. The section's count always shows the full total. Changing the filter clears your selection, and Select all means the rows currently shown.

Accounts nobody has signed in to, and deleting them together

Inactive: no sign-in for [30, 60, 90 or 180 days] narrows People to accounts whose last sign-in is at least that long ago, plus every account that has never signed in — somebody invited or created who never came. It starts on 90 days; choosing another number turns the filter on. Never-signed-in accounts come first, then the longest idle, and each row's line says last seen with the date, or never. It composes with every other filter, and the Inactive (no sign-in, 90 days) figure above the list is the same switch.

While it is on, the selection bar offers Select all inactive and Delete accounts…, which names the count once you tick some (Delete N accounts…). Select all takes the accounts shown but leaves out your own account, God Mode holders and system administrators, and the status line says who was left out and why; tick one of those on its own if you really mean it. Delete opens a confirmation that names how many, lists each account with its last sign-in, and says what happens to their data — the same words, and the same account delete, as Delete…. Type the phrase it shows (for example delete 3 accounts) to enable the button; Enter does not press it.

The accounts then go one at a time, with progress, and each line reports what happened: Deleted, or Not deleted with the database's own reason — for example the last Owner of an organization that still has members. A refusal never stops the rest. Your own account is never sent: the database refuses to delete the administrator who is signed in. The dialog cannot be closed while it runs, and the list is re-read when it finishes.

Not the same as Dormant, and not the same as an Inactive account

Dormant is account retention's own clock, which sends notices and may delete on a schedule. An Inactive account (the badge, and the Inactive users group at the end of the list) is one an administrator switched off. The Inactive: no sign-in for filter only narrows the list; nothing happens to an account until you delete it here.

Grouping the list by team

The People list is grouped by team, one heading per team, with anyone on no team gathered at the end. Click a heading to open or shut that group, or use the single fold button beside Group by team for all of them: while every group is open it reads Collapse all teams; once anything is shut it reads Expand all teams — its face always names the next press. Headings work from the keyboard too: tab to one and press Enter or Space. Your arrangement is remembered — which groups are shut, and whether you group at all — on this browser.

Grouping rearranges; it does not hide

Someone in two teams appears under both, and ticking them in one place ticks them in the other. Unlike the filters, turning grouping on or off keeps your selection, because no row left the list.

Sections
Overview Teams 2 Boards 5 People 12 System and Security Email configuration Human verification Storage Explorer Audit 18
Team All teams Is System Admin Search people… Clear filters Group by team
1 selected Select allDeselect all InactivateReactivate Delete…
John Doe System Admin 14 tasks · 2 teams · 3 boards Inactivate
Bob Smith 6 tasks · 1 team · 1 board Inactivate
Kate RossInactive 0 tasks · no team Reactivate

Account retention: notifying, then deleting, idle accounts

Inactivate and Delete are things you do to someone. Account retention is a schedule the product follows on its own for an account nobody has signed into in a long time. It ships off — nothing is notified or deleted anywhere until an administrator turns it on.

Three thresholds, set in Account retention settings:

First noticeDays since last sign-in before a first warning email. The account then shows as dormant in the People list.
Final noticeDays since last sign-in before a second email naming the deletion date. Always later than First notice.
Grace periodDays after the final notice before deletion can actually happen.

Read back as a sentence: an account idle for … days is warned, warned again at … days, and may be deleted … days after that.

One sign-in resets the whole clock

Idleness is measured from the last sign-in, recalculated every run. Signing back in at any point — including after a notice — clears any notices sent and scheduled deletion, and the count starts over. There is nothing to undo by hand; signing in is the undo.

A notice that never actually sent never leads to a deletion

An account is only deleted once both emails are confirmed sent, not merely scheduled. A bounced notice, or mail switched off system-wide when one was due, holds the account rather than deleting it on the strength of a message nobody received — it is flagged for review with the reason, rather than disappearing silently.

Two more controls travel with the feature: Dormant accounts, a filter beside Team and Is System Admin, showing accounts past their first notice; and Exempt from account retention, on an individual row, for a shared or service account or anyone who should never be swept up — it clears any scheduled deletion for them immediately and keeps them out of the idle count from then on. It asks why, and refuses an empty reason: “An exemption needs a reason.”

What is deleted follows the same "everything it owns" rule as a manual delete, with one difference: shared work is handed to someone else first. A task assigned to somebody else stays with them; a board or shared note goes to whoever on it has been there longest in the most senior role; an OKR hands to another contributor. A task they only held as assignee — never owned — is kept rather than left pointing at an account that no longer exists. A personal task goes back to its owner, because a personal task is never left unassigned; a task on a team board is kept and unassigned. Either way a comment from System says what happened — on a board task that it needs reassigning, without naming them; on a personal task that it went back to its owner, and when they left because their access to the organization was removed, it names them (by their name, never their email address). Inactivating someone touches no assignment — only the deletion path reassigns anything.

Deleted accounts is the record of every account this feature has removed, searchable by name or email, with what was transferred and to whom, and exportable to a spreadsheet. It is permanent and cannot be undone from here. It ages out on its own after a configurable number of months. Temporary accounts that STRIDE's own automated tests create and remove are not people, so they never appear here.

Teams and Boards

Two sections. Teams lists real teams with their people, admins, boards and groups. Boards lists every board with the team that owns it. A team can only be deleted outright once it is empty; if it still holds boards, people, groups or OKRs, the console shows you what and how many, and offers Inactivate instead — closes the team to its members, keeps everything, and can be undone.

Teams has three figures above the list: All teams, Active and Inactive. Press Active or Inactive to show only those teams; it sets the same status filter as the dropdown, so the two always agree. Press the lit figure again, or press All teams, to go back. The figures combine with search, clear any selection (so a bulk action can never reach a team you cannot see), and Select all takes only the teams shown. Nothing is remembered between visits.

A team that still holds anything can also be deleted all at once — and that can include accounts

When the plain delete refuses, the same dialog offers Delete everything in it — the team, its boards and tasks, its OKRs and check-ins, its groups and roster, counted by kind first. This is the same control a team administrator or Organization Owner can reach from that team's own Team settings; this console just reaches every team without opening each one. Keep the people's accounts is ticked by default — unticked, anyone whose only team and organization membership is this one has their account deleted outright, with everything they own. It refuses if the roster holds a peer Organization Owner, or if removing accounts would leave the organization with no other team.

Deleting a board deletes the team tasks on it, with their comments, watchers, reminders and links to OKRs (the OKRs themselves are untouched). Two kinds survive: anything an individual keeps on their own list rather than the board, and a team task that also sits on another board, which is simply taken off this one. The console shows you the counts first and asks for a typed confirmation. This cannot be undone — inactivate instead if you only want it out of the way.

Boards has the same controls as People. Both sections group under the organization each team belongs to — on Boards, organization then team then board. Group by organization and the fold button work both levels at once; a team or board belonging to no organization gathers under its own No organization heading. Which way you arrange it is remembered on this browser; the filter is not, so the list always opens showing everything. The same rules as People apply: changing the filter clears your selection, and Select all means the rows currently shown — acting on boards that scrolled out of view is the mistake worth preventing.

A team's row also offers a detail view — Overview, People, Boards and Activity as tabs of one dialog. It is read-only: its People on this team and Boards in this team buttons are doors to the filtered People/Boards view, and a third leaves for the team's own settings on the team board — never a second editor here. A team with none yet shows No boards in this team instead of the list.

Sections
Overview Teams 2 Boards 5 People 12 System and Security Email configuration Human verification Storage Explorer Audit 18
Team All teams Clear filters Group by organization Collapse all organizations
Meridian Group
Northwind Digital
Roadmap Northwind Digital · 4 members · 22 tasks Inactivate
Delivery Northwind Digital · 6 members · 41 tasks Inactivate
Campaigns Inactive Northwind Digital · 3 members Reactivate

OKRs

Between Boards and People: the one place an administrator can clear out objectives that should never have been kept — a duplicate import, a quarter of drafts nobody adopted, a team's whole set after a reorganisation. Objectives are grouped by team, and each shows its key-result and check-in counts before you decide anything.

In the summary, Candidates shows every objective and is lit while nothing narrows the list; with a filter on, it adds N shown underneath. Press it to clear every filter and your selection. The other figures, including Selected, only report a count.

It works at the objective level, never the key result — deleting an objective takes its key results, check-ins, comments and task links with it. If a key result alone is wrong, edit it on the OKRs page. Deleting a team's objectives asks you to type the team's name; the count is checked again on the server first. None of it can be undone.

Email configuration

Everything that decides whether a message leaves the building: the system switch, which teams may send, and a bench for trying one first.

Application email

Application email switches notification email on or off for the whole system. Off means no mentions, assignments, shares or ownership emails from any team — the Message Center carries on as is. Signing in, password resets and other account email are not affected by this switch and keep working regardless, so the switch can never lock anybody out of their own account. The password reset path moved onto the same mail system on 6 September 2026 but stayed deliberately exempt from this switch.

Beside it: whether the mail provider is configured, how many messages were sent, failed and suppressed today, and when the last one went. Suppressed is not a fault — it means delivered in the app and deliberately not emailed (a switch was off, it was the reader's own action, or there was no address). Send test email sends one message to your own signed-in address and nowhere else.

Teams

Which teams may send, and which events they send. A team ships off until turned on — here, or by a team admin in Team settings → Email; both write the same setting. Events opens the same seven choices the team sees. Inactive teams are listed rather than hidden — a team that is closed but still switched on is exactly the thing worth finding. While application email is off system-wide, no team can be switched on — the control is grayed out everywhere and the database refuses it either way; turning email off is never the direction that is blocked.

Email bench

A place to try an email before turning it on for anybody. Nothing on it changes a setting, and it can only send to your own address — there is no recipient field, deliberately. Preview shows the real template with sample content instead of anybody's actual task or comment. Send to me delivers it to your own address. It works even while the master switch above is off, limited to one per event per minute.

Preview as a folded email (3 sample updates) is a checkbox beside Preview. It is enabled for the six events that can fold, and shows the folded digest built from sample data instead of the single-event template. Ticking it previews and sends the folded sample: Send to me delivers the same folded digest the preview shows.

Trace answers would this actually send? — pick an event, team and person, and it walks the same checks the real sender does against the switches as they stand right now, reporting which one would stop it. It sends and records nothing, and never shows anybody's address. Two checks cannot be answered without a real event and say so rather than pretending: whether the recipient caused it themselves, and whether that exact notification has already been sent.

Sections
Overview Teams 2 Boards 5 People 12 System and Security Email configuration Human verification Storage Explorer Audit 18
Teams
Northwind DigitalSending Events
Field OperationsOff Events
Application email
On provider configured · 14 sent, 0 failed, 3 suppressed today Send test email
Email bench
Event · You were mentioned PreviewSend to me Trace
Trace: would not send — the team’s switch is off. Nothing was sent and nothing was recorded.

Email health

Window sets how far back this reads — Last 24 hours, Last 7 days or Last 30 days, the same three options the de-dupe and intelligence tab below offers. Five figures: delivered, failed, suppressed, failure rate, and when the last message went out. Suppressed is not failed — it is the system deciding not to send (a team switch is off, the reader caused it, no address on file); a failure is the system trying and not getting there, and the only figure worth acting on. The failure rate is counted against attempts only. Failures are listed first, suppressions after, each with the sender's own reason string.

Deliveries: what, where and how

One line per delivery attempt: who, the event, the address it was actually sent to, the team, status, reason if stopped, when, and the provider's message id (the string to search for in Resend). Filter by status and event, page with Show 60 more.

Retrying a failed delivery: select failures, and the button says how many it will attempt; it re-runs every check against the settings as they stand now, so anything a switch now stops is suppressed rather than sent. Each delivery shows only its latest state, but a retried row carries a ▶ disclosure with every earlier attempt. Retry queues (nothing sent yet, reversible with Cancel retry); Send queued, below the list, actually sends, at most 25 at a time.

Failures with no answer at all (a dropped connection, a 5xx) cannot be confirmed as failed — the message may have gone through, and resending can deliver a second copy. These are held back by default; the tick box beside Send queued includes them only after a second confirmation, and resets itself after every run. An unconfirmed badge marks a row where the send likely succeeded but the result was never recorded — treat it as delivered unless you know otherwise.

Only a genuine failure can be selected. A delivered message has nothing to retry; a suppressed one must never be resent, since the system already decided the recipient may not see it. Password resets, sign-up codes and address-verification codes are not retryable either — each is a one-time code with its own expiry, so use that account's own Forgot password or Resend code. Task reminders, test/bench sends and invitations are also excluded, each for its own reason shown on hover.

Outbound email de-dupe and intelligence

A second tab beside Outbound notifications. Notification email sits in a short queue before sending, so a burst of edits to one subject folds into one email. This tab configures that folding and shows what it has done — a health strip and the batches produced. It reads and configures; it does not purge. The settings switch consolidation on/off, how long a batch waits, and whether the model may additionally withhold an update it judges not materially new — both the master switch and withholding ask for confirmation, since withholding drops an update with no way to recover it. Withholding is decided for a whole email, never update by update. When Act on suppression is on, which it is in production, a folded email the AI judges not materially new against what the reader was already told is not sent, and its rows show Suppressed (NOT_SIGNIFICANT). An email containing anything addressed to the reader (a mention, an assignment, a note share or OKR ownership) is never withheld, whatever the verdict. With the switch off, such an email is sent and the row shows NOT_SIGNIFICANT_SEND_ANYWAY. The list of updates in a folded email never depends on the AI; a model failure removes only the ‘In short’ line.

Pending: what is about to send

The health strip's Queued now figure is a control, not just a number: it is a button: press it and the page scrolls to the Pending list — notifications still waiting out their grace period, one line per person and subject. An empty list is the common case — healthy, not a broken page. A group holding more than one update discloses its individual queued items, oldest first. Names shown are display names, never email addresses, and nothing here shows a notification's content. While a group's items are still being fetched it shows Reading the pending updates… in place of the list.

Inbound Tasks by Email

A third tab: the application-wide switch and roster for turning an email into a task. Five parts — which organizations and teams may use it, a masked audit of processed messages, the service switch and health, held mail, and a bench scoped to your own address. Nothing shows a message body, a signed URL or any credential.

Reconcile warning. The service health tiles show how many messages are waiting, stuck or failed, and when the reconciler last ran. If the reconciler hit something worth knowing, such as the mail provider reporting more messages than it returns at once, a yellow Reconcile warning note appears under the tiles saying what happened. It disappears once a later run has nothing to report.

Held mail. A message from an address that belongs to a STRIDE member, but that the mail provider cannot confirm really came from that sender, is held instead of becoming a task. It makes no task, uses nobody’s AI key and shows in nobody’s list, including the member it names. Each held message shows its domain, sender, subject (marked unverified), why it was held and when it expires. Choose Allow once to process just that message, Always allow this domain to accept every address at that domain from now on (this also releases its other held messages), or Discard. A message nobody decides is dropped after 14 days and stays in the audit as expired. Mail from someone with no STRIDE account is never held: it is refused as before.

Allowed domains. Domains are matched exactly, so allowing example.com does not allow mail.example.com. An allowed domain lets through only mail from a domain that publishes no policy of its own; a message whose check could not run stays held, and you can allow just that message once. A message that failed its domain’s own published policy is a forgery signal and is held even for an allowed domain, and Always allow is disabled on it with the reason shown. Free mail services such as Gmail or Outlook cannot be allowed, because that would admit anyone with an account there; genuine mail from them verifies anyway. The cost of allowing a domain that publishes no policy is that a forged address there cannot be told from a real one; each use is logged.

Verification mode and alerts. Hold unverified (the default) holds unverified mail, Record only writes the check result to the audit and holds nothing, and Off skips the check. If every message is suddenly held, switch to Record only and allow them while you look. Every God Mode holder is also emailed about each held message, up to a per-hour limit you can set, then one summary for the rest of that hour. Only God Mode holders can see or change any of this. While the lists load the panel says “Loading held messages…” and “Loading allowed domains…”; with no domains allowed it says “No allowed domains. Every message must pass verification.”

Replies to a task’s thread

This switch decides what happens when an email continues a conversation that already became a task.

Changing the switch is recorded in the review log. Matching uses the earlier email’s subject and wording within 30 days, on the recipient’s own tasks only. It does not use mail headers yet.

System AI: the key for STRIDE’s background jobs

Configuration has a System AI pane directly under Email configuration. Only God Mode sees it, and only with a confirmed sign-in. It chooses the key that STRIDE’s own background jobs use: combining notification emails into one, and folding duplicate Message Center updates. Nobody’s personal key is ever used for them.

There are three sources, chosen with Save source:

With a system key set, its card shows only the provider and the last four characters of the key, and who set it and when. The key is never shown again, and nothing on this page can read it back. Replace key takes a new one, tested the same way before it is kept. Test makes one small call and says whether the provider accepts the key now. Remove… asks first (Remove the STRIDE system key?) and deletes the key; the jobs then skip their AI step until a key is set again.

Daily limit is the most calls the jobs may make in one UTC day: 200 unless changed, 1 to 5000. At the limit the jobs skip the AI step until midnight UTC, and the pane says they are paused. Pressing Test does not count toward it.

What the jobs did counts the last 24 hours: calls, failures and when the last one was made. When the jobs are not calling the model, the line says why in plain words, for example Skipped: no STRIDE system key is set. A skip is not a fault; the emails still go out.

Needs attention appears on the pane, and as a chip on Overview, when the key has been unusable for more than an hour: the provider rejected it, it is out of credit, or it cannot be read. Set it again, or choose another source.

Where the calls appear. An organization’s key puts them in that organization’s AI insights and Datadog. A system key puts them in no organization’s. The organization’s owner is told, on their AI card, that STRIDE also uses the key for these two jobs and, when those jobs include updates for people in other organizations, how many. The owner sees the calls, never their content. Every change to the source or the key is recorded in Audit, with the provider and the last four characters, never the key.

Ask STRIDE before sign-in is a second card in the same pane, for God Mode only. It decides whether the panel opened from the sign-in page’s own Ask STRIDE control answers anybody, and whether a new Organization owner with no AI key gets starter answers while setting STRIDE up. It starts Off. A chip on the card shows Off, On or Stopped for today. Answers come from STRIDE’s public help only, after a security check, and are paid for by the System AI key chosen above. They appear in that organization’s AI insights but have their own daily question limit and never use the key’s daily call limit, so visitors can never stop STRIDE’s own AI jobs.

Ask STRIDE for readers without a key is a third card in the same pane, for God Mode only. It decides whether a signed-in reader with no AI key anywhere, none of their own and none set by their organization, gets full Ask STRIDE answers on STRIDE’s shared assistant while they set up their own. It starts Off. A chip shows Off, On, Stopped for today or Not answering (the System AI key would not answer now). Answers are paid for by the System AI key chosen above and recorded against its operator, never the reader, with their own question budget, so they never use the key’s daily call limit. A reader whose own key fails is never moved onto it. The reader-facing side is in Ask STRIDE with no AI key.

One more background job runs outside STRIDE and is not set here: the weekly drafting of help-feedback suggestions runs in GitHub on the organization’s own key, held as a GitHub repository secret.

System and Security

Two-step sign-in, above password strength, makes everybody confirm a 6-digit code emailed to their sign-in address the first time they sign in on a browser, and again after Ask again every business days (1–30, 7 to begin with; Monday to Friday, UTC, holidays not skipped). It is off by default, so turning it on is your decision. The switch and the number save together with Save two-step sign-in. A notice beside the switch warns that saving it on makes your own session unconfirmed: you are asked for a code emailed to your sign-in address straight away, so be ready to open that mailbox. Turning it off means a password alone opens every account, so an amber note appears and every System Administrator is emailed. Turning it on is refused while an administrator has no proven address.

Password strength — Medium, High or Extra-High, each listing what it requires. Set the matching minimum in the Supabase dashboard too: the app states the rule, only the service enforces it for every route in.

Attachment rules

The size limit is set here, in megabytes. It starts at 5 MB and can be raised to 25 MB at most, and it holds however the upload is made. Files already attached are never affected — the limit only applies on upload.

The other protection is the list of accepted file types, and it works the way round that matters: a list of what is allowed, not a list of what is banned. A type not on it is refused, and told to put the file in a .zip instead, which always works. The list is fixed, not configurable. It covers documents, spreadsheets, presentations and their Apple/OpenDocument equivalents, images, audio/video, archives, mail/calendar/contact files, and common project, CAD and design formats. It deliberately excludes anything a browser renders as a page from a signed link (.html, .svg, .xml), anything a runtime executes, and any installer or disk image; a file with no extension is refused too.

The list in the browser is a convenience, not the control. The size rule that actually holds is in the database.

There is no malware scanning. Uploads go through one file service that checks the size and the type before it stores anything, and nothing inspects what is inside a file.

Orphaned data moved to the Storage Explorer, where the rest of space accounting lives.

Storage (Cloudflare R2)

Where screenshots and new attachments are kept. Only a God Mode holder can change it, and only in a confirmed MFA session. The secret access key is stored encrypted, and it is never shown again once saved: the field shows that a value is set, and Replace is the only way to change it.

The panel loads what is already in use. A line at the top says where the settings come from: In use is shown as a chip beside the label, reading This panel, Environment secrets or Not configured. Account ID and Bucket are filled in for you; the ID is shown in a monospace face on one line. A key is never shown. Each key field reads Not set, Set (saved) · ends …XXXX or Set (environment) · ends …XXXX, where XXXX is the last four characters (shown only for a long key). Enter sets a new value; Replace changes an existing one. Labels, inputs and buttons line up on one left edge, and on a narrow window each label sits above its field. Reading these settings needs a God Mode holder with a confirmed second step (MFA); anyone else sees the panel refuse.

Test connection checks the endpoint, the credentials, the bucket, a write and a delete, and a read back from the browser. Save runs the same test first and refuses any value that fails it, so a saved configuration has passed. Every change is recorded in the storage settings audit.

System Admins and God Mode

Three roles share the word "admin" and are not the same thing. Team admin — authority inside one team or board: its people, groups, swimlanes and topics; one exception, anybody who administers a team reads the whole organization's roster, though assigning people across the organization stays with an Organization Owner. God Mode — authority over the whole application: every section of this console, every account, every team's data, the platform settings, and the ability to permanently purge content and files. System Admin — authority over one organization from this console: its people, teams, boards, OKRs and role changes, and nothing else.

Only a God Mode holder can make a System Admin, or another God Mode holder; a System Admin cannot. See the People section above for God Mode, and for a System Admin open People → Profile → System Admin, choose the organization and press Make System Admin; Remove System Admin takes the role away. Only the person's one active organization is offered. A person who is active in several organizations, or who holds God Mode, cannot be made one, and the database says so. Making someone a God Mode holder ends their System Admin role in the same step, and removing God Mode later does not bring it back.

A System Admin cannot see Distributed builders, System AI, API bench or any platform setting, cannot delete an account, team or board, and cannot deactivate, reset the password of, or change the role of a God Mode holder or another System Admin. Opening a God Mode holder’s Profile shows a System Admin their name, role and organization counts only: not the person’s alternate email, mobile number, API key or note count. The role ends by itself, and is recorded in Audit, when the person leaves the organization.

This section lists who holds a system role and every grant or revoke, with who made it.

Three things the system will not let you do

You cannot change your own role — ask another God Mode holder. You cannot remove the last God Mode holder — grant it to someone else first. Hiding the button is not the rule — all three checks live in the database. There is no such floor for System Admin: an installation with none is valid.

Revoking is immediate and narrow: the person loses this console straight away, and their account, teams, boards and tasks are untouched.

List paging, below the attachment size limit, sets two numbers for every organization at once: Rows per request (50–1000, 500 to begin with) is how many rows each request fetches from the database, and Show older loads (25–500, 100 to begin with) is how many older completed tasks one press of Show older adds. One Save paging saves both, and it takes effect the moment it is saved. Every list still loads in full — this changes how it is fetched, not what it holds.

Sections
Overview Teams 2 Boards 5 People 12 System and Security Email configuration Human verification Storage Explorer Audit 18
Password strength
Medium
At least 10 characters, a letter and a number.
High
12+ characters, mixed case, a number and a symbol.
Extra-High
16+ characters, no sequences, not your email name.
Attachment size limit
Largest attachment, in MB Save limit 1–50. Files already attached are never affected.
List paging
Rows per request, and Show older loads Save paging Applies to every organization.
System administrators
John DoeSystem Admin granted 4 Aug by Bob Smith

Human verification

Decides whether a stranger must prove they are a person before the application will register an account, start a password reset, or keep accepting failed sign-ins. It uses Cloudflare Turnstile and applies to the whole application. It is off until you switch it on, and where it stands is reported on Overview, directly below Password strength.

What it is, and what it is not

It slows down automated abuse. It does not decide who anybody is. Passing the check creates no account, proves no email, grants no permission and signs nobody in. It is added to the codes this product already emails, never instead of them.

The controls

Signing in normally is never challenged — only repeated failures bring the check out.

The secret key is written, never shown

The secret field is write-only, and that is deliberate: you can set it or replace it, but you can never read it back. Nothing in this console can display, echo or partially reveal the value — the database answers only yes or no to whether one is present. If lost, roll a new one in Cloudflare.

Saving a secret here does not by itself switch verification over to it

The value the server actually checks against is set where the application's email function keeps its configuration, not in this box. Saving here records it safely; making it the value in force is a deployment step your engineer performs. If verification is on and readers are getting through after a few tries, an unset or mismatched secret on the server is the first thing to check.

Turning it off is safe

The master switch is the whole rollback — switching it off returns every flow to what it did before, with no account, membership or password changed either way. If Cloudflare itself is having an outage, the product already handles it: readers are refused a few times and then allowed through on the emailed code instead. There is deliberately no per-person exception — if one reader's browser will not draw the check (usually an ad blocker or a locked-down network), see The security check in the sign-in manual for their fixes, or switch the feature off for the whole team while it is sorted out.

API bench: trying the Task API

The Task API lets another system create, read, update and delete a person’s tasks. API bench sends those calls for you, the way an outside system would, and shows what that system would see: the status, the request and the response. Only God Mode holders can see it. It is the last section under Configuration, after Human verification. It reads the Task API’s counts when you open it and nothing else until you press Run.

The Task API: on or off, and how much it is used

The top band, Task API, says whether the API is On or Off. It ships off. While it is off it refuses every token and nobody can create one; the dry runs in the bench use your own session, so they work either way. Switch on and Switch off each ask you to confirm first (Switch the Task API on? and Switch the Task API off?); switching on lets every signed-in person who is not an administrator create tokens in My Settings, and switching off refuses every API-token call at once, without deleting anything. Each switch and each revoke is recorded with who did it, in the Activity Log of User traffic and audit trail. Refresh reads the figures again.

Trying a call

  1. Choose the kind of task and the operation. Individual task or Board task, then Create, Read, List, Update or Delete. A board task asks for the Board, from the boards you are a member of; it lands in No topic, and its status is one of that board’s lanes.
  2. Fill in the fields. A field left empty is not sent, so an update changes only what you filled in. Read, Update and Delete need a task id: run a List first and the first task fills in the id and its version for you. Update needs the version the task last showed; leave it empty to see the API refuse an update that does not say which version it was based on. On Create, If the title is already open picks what a clash does: Merge into it (default), Refuse with a 409 or create another. The completion choice on Update reads Leave as it is, Mark complete or Reopen.
  3. Edit as JSON shows the request body as text. When it holds JSON, that is sent instead of the fields, which is how you try a field the form cannot send, such as an unknown one. Text that is not valid JSON sends nothing and says so.
  4. Press Run. You see a status such as 200 OK or 409 Conflict, how long it took, one sentence on what the answer means, the Request as an outside system would send it, and the Response. Copy puts the whole response on the clipboard (it says Couldn’t copy if the browser refuses). After a List with more tasks, Next page asks for the next page with the cursor the last one returned.

Dry run as me saves nothing

Dry run as me is the only way to try a call. It runs each call with your own permissions through your own signed-in session, and a write is performed and then discarded, so nothing is saved and the sentence under the status says so. A task a dry Create would have made never exists, so its id is not offered to the next operation. Read and List only read what you could already see. The bench never writes your own tasks. No token is involved: the request block says the token was not sent.

A refusal looks like the API’s

A refused call shows the status and the code an outside system would receive, and a plain sentence. A clash with an open task of the same title is a 409 in the amber status, and the response carries the id of the task it collided with; a change made since you read the task is a 409 with its current version; an invalid field is a 422; a task you cannot see is a 404; a call refused as the app would refuse it is a 403. A 401 means this browser has no confirmed sign-in: sign in again and enter the emailed code. An unexpected database error is shown as a generic refusal and never as the database’s own words.

Storage Explorer

Answers three questions: where is the space, what should age out, what needs attention and what can come back. Visible only to a God Mode holder.

Looking around is safe. Only one path deletes anything.

Every card, chart, chip and row here filters and navigates. None of them removes anything. The only way to delete content is Retention → Preview → type the confirmation, and you can back out at any point before that. A restore only adds rows.

Four views, and what each one answers

Switching between them keeps your selection. Refresh re-reads everything; the timestamp beside it says when figures were measured.

Explorer: where the space is

Four summary cards at the top — Total storage, Purge opportunity (what could safely go), Cleanup & orphaned data, and Database health — each a door to the view that can act on it.

Screenshots and Attachments cards

The Screenshots card is an exact figure: the bytes and the number of files held in R2. The Attachments card is exact too, and it counts both the files in the older attachment store and the new files in R2, so a figure does not change when a file is moved.

One filter, and every chart obeys it

A single filter — team, data type, age, state, plus search — governs the whole view. No chart keeps a filter of its own, so nothing on screen can describe a different selection from the thing next to it. Set it from the dropdowns, or by selecting a visual (a treemap tile, an age bar, a type slice). Active filters appear as chips, and each chip can be removed on its own; Clear filters removes them all. A figure that deliberately does not move says application-wide; one that responds to your filter is labeled Filtered footprint.

The state filter is a lens, not a promise: Purge eligible and Protected are deliberately cautious estimates. The numbers to act on — on the cards, in Retention Opportunity, in the simulator and the preview — come from the purge planner itself.

Sections
Overview Teams 2 Boards 5 People 12 System and Security Email configuration Human verification Storage Explorer Audit 18
Explorer Retention Maintenance Refresh
16.0 MB
Total storage · application-wide
1.2 MB
Purge opportunity
0
Cleanup & orphaned data
Lean
Database health
Team All teams Data type Attachments Age 365+ days Clear filters
Attachments × 365+ days × Filtered footprint · Attachments › 365+ days
Storage by team and data type
Northwind Digital 4.2 MB · tasks, comments, attachments TreemapBarTable
No Team / Personal auto: off 88 KB · notes are never purged

Where to look first

Storage by team and data type as a treemap, bar chart or table — the same numbers three ways. Every consumer ranked, under the Explorer, lists teams and people together; personal rows show a name, count and size, never content. Each row with something removable offers Preview, which selects that scope and fills in the age — it does not delete anything. Personal tasks are a single scope for every account, so Preview on one person's row selects all personal tasks, not just theirs — the console says so when pressed. Largest attachments lists the biggest files, marking those on completed work.

Looking inside one team or scope

Selecting a team opens a panel with its overview (by tasks, comments, objectives, key results, insights, attachments), retention state, files and history. Its two buttons are doors, not a second set of controls: Preview purge opens the preview with this scope selected, Edit policy opens the one policy form — there is deliberately no second policy editor in the panel.

Every task is counted, and where it is counted is not who owns it

Every task is counted, not only board tasks. An individual task is reported under its owner's default team, or their first team. Tasks counted says how many of the total are accounted for.

Where a task is reported is not who owns it

This grouping only makes the footprint add up. It gives nobody access to anybody's task, and it never decides what a purge deletes — a retention policy only removes content that reaches the team through a board it is actually on.

No Team / Personal

No Team / Personal is a real scope: it is not an error, a missing team or a gap in the data. It holds tasks belonging to people who are in no team, and everybody's personal notes and comments on them. Personal notes are counted separately and are never purged — no policy anywhere can remove one. The same name selects the personal retention policy in the Retention view; see Personal tasks below.

Two kinds of number, and they are not the same promise

The console marks every figure as exact or estimated and never adds the two into one "you will save X".

Retention: what should be allowed to age out

Works on one scope at a time — a team, or No Team / Personal — chosen at the top, shared with the Explorer's team filter and the panel.

Trying a cutoff before you commit to one

Type a number of days, or use the 30/60/90/180/365 shortcuts, and the answer updates live: records, space, and how far back it reaches. It deletes nothing. The figures come from the same planner a purge uses. Retention Opportunity shows what the current settings would actually free, with a Preview purge button.

What this applies to

Anything that includes open work warns you at the point you choose it, again in the preview, and again at confirmation — age never makes live work safe to delete. You can also choose which kinds of content the policy covers — tasks, objectives, key results and attachments.

Screenshot cleanup: preview, then a typed count

Retention also removes screenshots that no note, task or comment still shows, and screenshots that belong to something already deleted. The age is at least 7 days, and a shorter one cannot be chosen.

Preview cleanup works out what would go and writes nothing. Nothing is deleted until you have previewed the plan and typed the number of screenshots shown. If the set of screenshots has changed since the preview, nothing is deleted and you are asked to preview again.

Automatic retention

Every team has a policy from creation, with conservative defaults: closed content older than 180 days, evaluated daily. The form is plain language: on/off, age, a Check for cleanup frequency (Daily, Weekly or Monthly), and Protect active work. One policy per scope, one form that edits it.

Open work is never purged automatically

The default scope is closed content only. An automatic run will not touch an open task, objective or key result unless you deliberately change the scope.

It runs in the database, on a schedule — you do not need the page open.

How you know it saved

Save retention policy confirms itself beside the button, and it reads the stored policy back to you rather than repeating what you typed — e.g. Retention policy saved at 2:32 PM — on · daily · keep 180 days · closed content only · tasks, attachments. If the panel cannot read the policy back it says so instead of claiming a clean save. Every other action on this page that changes something confirms itself the same way. A policy with no kind of content ticked would be switched on and unable to remove anything, so it is refused rather than saved.

Refreshing the page reopens the same scope you were last viewing. If nothing has been chosen yet, the panel says No scope selected rather than showing an empty form.

If the figures cannot be read

A message across the top means the storage figures could not be fetched — usually a session that has ended; reload to sign in again. While that message is up, the numbers are withdrawn rather than left on screen.

Personal tasks

Personal tasks means every task that sits on no team board, whoever owns it. It has a policy like any team. Every task belongs to exactly one of the two scopes — a team's board, or personal — with no overlap.

This one is switched off by default, on purpose

Every team's policy arrives on. This one does not — these are somebody's private tasks, and nobody administers them the way a team administers its board. Turning it on is your decision, not a default.

Personal notes are never purged, anywhere. Only tasks, their comments and attachments are in scope — no objectives or key results exist on a boardless task. Age is measured from the completion date for finished work, not creation. Because the content is personal, the preview names the owner of every task it lists and how many accounts it affects, and the warning before confirmation is the strongest in this console.

The Lean storage advisor

Measures what several retention cutoffs would actually free, across every team and the personal scope, and recommends one — arithmetic over your real data, from the same planner the purge uses. It never deletes anything; its buttons fill an age into the policy form or open the preview. It does not simply recommend the shortest cutoff (that is just "delete more") — it picks the gentlest cutoff that still captures at least 80% of what the shortest would, and shows every cutoff it measured. If nothing has aged out yet, it says so and offers no button.

Purging now

Choose scope, age, what it applies to and the content kinds, then Preview…. Always preview — it is the only route to a deletion. The preview states outright that nothing has been deleted, names what would go (tasks, objectives, comments, mentions, check-ins, insights and links), the date range, the space it would reclaim, and what it skipped and why. The purge deletes the plan you were shown, or nothing — if the data changes between preview and press, it refuses and asks you to preview again.

Type it to confirm — it is restorable, but only from the Recovery archive

The button stays disabled until you type purge now. Every row a purge removes is captured in the Recovery archive, restorable by a God Mode holder for 12 months (how); attached files are not kept yet (file not kept). If the Recovery archive is not set up, only a database backup or the provider's daily snapshot can recover it. Open work in the selection makes the warning louder.

Changing any filter clears the preview.

Maintenance: what needs attention

Everything here is a reading or a repair; nothing purges content.

What the space is made of

Row content per data class, beside physical relation and index bytes — expect the two to disagree. Some of the largest classes are never purgeable (profiles, personal notes). The age table splits tasks by age, team vs. personal, open vs. completed — open rows are shown in a quieter style because they are protected — no policy removes them unless you deliberately choose the open scope. Database health is one line — lean, healthy, growing or needs attention — not a percentage of plan, since Supabase does not tell the application what your limit is. A stuck file or broken reference outranks size. Maintenance state reports live and dead rows per table — nothing on this page runs a VACUUM.

Removing queued files

Attachments and screenshots live in file storage, which the database cannot delete on its own, so a purge queues files and they are removed afterwards. Queued files in R2 are removed automatically every hour, with no one needing to press anything; the hourly job can only remove files that are already queued. You can still remove them at once with Remove queued files on Maintenance, which also drains the older attachment store (the hourly job does not). Until gone, the run is reported as partial; anything unconfirmed stays queued and is retried on the next hour or the next press.

Deleting a task queues its attachments and screenshots the same way, however it was deleted: by its owner, by an import undo, by a purge, or with an account or a team. They show as purge files still queued until they are removed, normally within the hour. Archived tasks keep their files. Deleting a note queues its screenshots in the same way, and deleting a team queues the screenshots pasted into its OKR comments; both are removed within the hour like the rest.

Database snapshots

Under Database on Maintenance you can take a manual copy of the whole database. Press Take snapshot…, give it an optional label and note, and confirm. The label defaults to manual. The copy runs in the background; the list follows the snapshot you started, showing its status until it finishes. A snapshot someone else took appears when you press Refresh.

A snapshot is also taken automatically once a week, early on Sunday (UTC). It is labeled weekly, and the list shows Weekly schedule as who asked for it. A week that already has a weekly snapshot is left alone, and if a manual snapshot is still running at that moment the weekly one waits for the next week (or for the owner to run it again by hand).

Each snapshot is encrypted before it leaves the process that makes it. Download gives you the encrypted file through a link that works for 60 seconds. Nothing on this page can open it: only the offline key held by the owner of the deployment can. Restoring a snapshot is done by a written runbook, never from this page.

Snapshots include user accounts and their sign-in methods, and exclude active sessions. Anyone who can download a snapshot and open it holds the users’ account data, so treat the file as sensitive.

Only God Mode holders who have confirmed their second factor can take, download or delete a snapshot. Snapshots are deleted automatically 90 days after being taken, except the four newest, and the first one of each month, which is kept for 12 months. The list then shows them as Deleted. To delete one sooner, press Delete… and type its label. Deleting cannot be undone. When you hold 20 snapshots or 1 GB, the list shows a warning so you can clear old ones. If the snapshot job is not set up yet, the page says so and nothing is started.

Loose ends

Three kinds, kept apart on purpose:

Two proven kinds are deliberately left alone: a board whose team record has gone missing is a fault to investigate, not a tidy-up; a team with nobody in it is not a loose end at all — nothing in this application deletes a team on its own.

Purge history

Every run, manual or scheduled, with what it removed, skipped and failed — counts and sizes only, never the deleted content itself.

User traffic and audit trail

Who is using the application, when, and what they did — by person, team and feature. It sits above Audit: this is the operational record of product use, Audit is the administrative record of what an administrator changed. Nothing here changes anything — it reads, and that is all.

It records from the day it was switched on

The application kept no activity record before this section existed, so a date range reaching back before that point shows nothing for the earlier part — not because nothing happened, but because nothing was written down. The section says so rather than drawing an empty chart. A figure this application cannot honestly answer is absent, never estimated.

Five views

The date range governs the whole section

Defaults to the last 30 days, shows resolved dates, and applies across all five views. Clear filters keeps the date range — it is the period you are studying, not a filter you left on by accident. The data reaches back 90 days; older activity is in the Recovery archive’s Activity history.

Filters compose, and they are shared

Team, person, activity, feature and result all narrow the section at once, each as a removable chip; clicking a chart segment or a row applies the matching filter directly. A filter the recorded data cannot support is shown disabled with the reason, not hidden. The search box matches four things at once: the name of the person who acted, the feature, the action, and the label of the thing acted on — so a colleague's name finds their activity even on a row with no label, such as a sign-in, which is the case the box used to miss. It composes with every other filter rather than replacing them.

Signals are observations, not accusations

The Security view reports things like repeated failed sign-ins from written, deterministic rules, with no score, no grade and no risk rating — everything is an observation for a person to interpret. The list is paged — First, Previous, Next and Last, with a line saying which rows you are looking at out of how many; you choose how many rows a page holds, and changing that returns you to the first page. A separate search box above the list itself narrows only this list, matching person, event and detail text; it composes with the four signal cards, which keep counting every row for their rule regardless of the search. A search matching nothing is worded differently — the other means you have hidden it — from a range holding no signals at all.

What is kept, and what is deliberately not

Who, when, which team, which feature, what kind of action, success or failure, and a short reference to the thing acted on. It does not keep the contents of tasks, notes or comments, or any password, token or key. Access is restricted to System Administrators in the database itself, not merely by hiding this section.

Exporting

Export takes your current filtered view or the whole date range, as CSV or JSON, with a choice of columns — it can only export what you can already see.

Audit: what has changed, and what is not recorded

Everything the application writes down about its own administration, newest first. Metadata only. What appears:

Filter by kind, search the text, press Show more to lengthen the list. Nothing here can be changed or removed.

Sections
Overview Teams 2 Boards 5 People 12 System and Security Email configuration Human verification Storage Explorer Audit 18
Kind All kinds Search…
Role System Admin granted to Kate Ross by John Doe 4 Sep, 09:12
Retention Scheduled run removed 214 rows, 2 files 4 Sep, 03:00
People Account created for Sam Okafor 2 Sep, 16:40
Show more

What nothing records

A number of administrative changes leave no trace at all, so their absence from the list tells you nothing either way:

This is not a compliance audit trail. An empty Audit list means "nothing of the recorded kinds has happened," not "nothing has happened." There is no export; select rows on screen and copy them.

Browser errors: what failed in readers' browsers

The failures readers' browsers reported on their own, with nobody filing anything: an error nothing caught, a background request that failed with nobody waiting for it, or a part of the app that did not load. Only God Mode holders can see it.

Choose a Window of 24 hours, 7 days or 30 days, and press Refresh to read again. Nothing here can be changed.

What leaves the browser is scrubbed first. Email addresses, quoted text such as task or note titles, tokens and the query strings of links are removed before a report is sent, and every record is deleted automatically after 30 days.

Feature usage: which features people use

Every control in STRIDE — a button, a tab, a menu row — is a feature, and this section shows how many people pressed it. It exists so that a feature nobody uses can be noticed and, later, retired, and a feature people miss can be made easier to find. Only God Mode holders can see it. It counts presses; it does not record what anyone wrote, any title, any link or any page address, and it never shows who pressed anything. Who pressed a control is used only to count distinct people. Everything a person did is kept for 90 days with their account, and then only anonymous monthly totals remain, for 25 months. What STRIDE counts about how you use it says the same thing to everyone else.

Reading the list

Press a row to open the detail beside it: its weekly figures, where it lives and any decision. The “…” on a row opens Actions for that feature: Open in page goes to the place in STRIDE where it lives, Read the help opens its manual page, Mark as records a decision, and Clear decision removes it.

Before sign-in: the sign-in page’s Ask STRIDE

On All features, above the list, a line reading Before sign-in: Ask STRIDE opens the totals for the Ask STRIDE panel on the sign-in page, where nobody is signed in and nothing else can be counted. It shows, for the Window you chose, how many questions were asked, how many were answered and how many were refused, the refused ones by reason (for example the security check failed, the rate limit or the daily spend limit was reached, or the question was off topic), and a day-by-day table of the latest days.

These are totals only. No question, no answer, no address and no person is recorded here, and the numbers are never tied to a team or an organization, so only the Window changes them; the other filters and Include STRIDE support do not. They are not part of the figures across the top and never make a feature unused. The totals are kept for 400 days. If Ask STRIDE before sign-in has never been switched on, the line says none yet.

When a feature may be called “unused”

Not early. A feature is called unused only after it has been listed for 30 days, over a window of 30 days or more, on a page that at least 10 people used (3 people for an administrator-only control). Until then the row says Not enough data, and a few controls are marked Rare by design (the ones you press once in a long while, like downloading your data) and are never called unused. While counting is under 14 days old the section says it is collecting and shows only People and Uses.

Decisions: recording, never deleting

Each decision is one of Watching (with a date to look again), Retire candidate, Make easier to find, Keep or Retired. A decision changes nothing in STRIDE. Retiring a feature is a separate piece of work. Marking a feature a retire candidate that the rule above cannot yet judge asks you to confirm and says which condition is not met. A decision needs a confirmed sign-in (the second step), and every change is recorded in the Activity Log of User traffic and audit trail: the detail’s Open the audit trail goes there. From User traffic, a link at the foot reads For which features people use, see Feature usage.

Review with AI

On Recommendations, Review with AI asks your AI Assistant to read the counts and suggest features to retire, to make easier to find, to merge into another, or to keep, each with a reason and, for a feature that deserves more use, an idea for where to place it. It uses your organization’s AI key, or your own; with neither it says your assistant is not set up and points to My Settings, and sends nothing. The AI capabilities list calls it Review which features people use.

Distributed builders

Distributed builders, the section that shows the machines that build STRIDE, has its own page: Distributed builders. It covers the machines and their work, answering a builder’s question, blocked or failed work and talking to the builders.

Onboarding someone

There is nothing to issue first. Creating an account is open, so the sequence starts with the team, not this console.

  1. Add them to a team: Team → Team settings → People. This emails them their invitation straight away, from that team's own name and mark, with a link that opens the sign-in page with their address filled in — there is no code in it.
  2. They follow the link and create their account with their real first and last name — shared tasks reach a person's own list by matching it exactly.
  3. They are emailed a six-digit code to prove the address is theirs. It lasts an hour; asking again replaces it. An account stuck here looks like someone who has not registered yet. See Verifying your email address.
  4. Give them a board — directly, or via a group that already has access. Team membership alone gives no board access.

You can add the address before they register; the two connect once it is proved. If an invitation never arrived, the team's Email invitations dialog sends it again at no cost.

Offboarding someone

Two options. There is also a third, automatic path for someone who simply stops signing in — see Account retention in People, above.

InactivateCannot sign in, reaches no data. Everything returns intact if reactivated. Use for leave, a role change, or anything reversible.
DeletePermanent. Removes the account and everything it owns — tasks, notes, tags, comments, memberships, uploaded files, and any board they created that nobody else is on.

Their tasks go with them

A person's tasks belong to them even when shared to a board, so deleting the account removes those tasks from every board they were on. If the work needs to survive, have someone else recreate it, or inactivate instead. An existing session can remain valid briefly after inactivation — for an urgent removal, also revoke sessions in the Supabase dashboard.

Who can see what

God Mode holderThis whole console, for every organization. Granted by another God Mode holder; the first is set up outside the app.
System AdminThis console for one organization: its people, teams, boards, OKRs and role changes. Made by a God Mode holder.
Organization OwnerThe organization: name and mark, people, invitations, licenses, and assigning anybody to any team, board or group inside it. A different axis from Super Team Admin, not a step above it.
Super Team AdminEverything a team administrator does, plus six elevated actions: creating a team, creating a board, configuring the team's email, importing a file of OKRs, sending the team's invitations, and granting this role.
Team administratorManages a team's people, groups and existing boards, and administers every board in that team. Creating a board is not theirs. Reads the whole organization's roster.
Board administratorManages one board's roster, columns, topics and tags — a separate axis from team role.
MemberSees the boards they were added to. On a board they belong to, they can edit any task on it — not only their own.

Someone can hold different roles on different teams. This table is a summary — the product holds the exact one: Permissions by role, opened from the button of that name in the Organization's People pane, lists every capability against all four roles and is generated from the rules the database actually enforces. See how to read it.

What this console cannot do

If something looks wrong

Run Storage Explorer → Maintenance → Orphaned data → Scan. It reports rows and files belonging to accounts that no longer exist, usually left by a deletion made outside this console. Scanning changes nothing; Clean up removes what it found.