Connector statuses
Last synced and item counts
Each connected tile shows when Hive last synced it and how much it has read, for example “Last synced 12m ago · 1,204 files · 312 emails”.- Last synced is when Hive last finished a clean sync round for that account. A round in which every read from the provider failed does not count, so the time keeps ageing while Hive cannot read the account. Before the first one, the tile reads Not synced yet.
- Emails are counted per connected mailbox (Gmail, Outlook) and summed on the tile. The Accounts panel shows each mailbox’s own count.
- Files are counted per provider (Google Drive, OneDrive, SharePoint and so on), not per account, and include only the files you could open in Ask yourself. Admins see their own count, not the whole workspace’s.
Reading accounts on the hourly pass
Hive reads every member’s connected accounts on its hourly pass.- Each workspace has its own minute. The hourly pass starts at a fixed minute in the first ten minutes past the hour, the same minute every hour for a given workspace, rather than for every workspace at once. The countdown at the top of the Signals page and on the Connectors tab shows your workspace’s own start.
- A provider hiccup waits for the next pass. When Nango or Hive’s sign-in service is briefly rate limited or unavailable, the account is not marked broken: that member’s read waits for the next hourly pass, and the other members’ reads carry on.
- A busy provider waits, it does not fail. Hive shares each provider’s read capacity fairly between workspaces (see Provider read limits). When a provider is busy across Hive, an account’s read waits for the next hourly pass. This is not a failure: the account is not marked broken, nothing is skipped (the read resumes from where it left off), and Signals does not count the account as read until the read actually happens. A first Signals read of a busy finance or CRM tool simply starts on a later pass. For Outlook, your stored email is still read that pass; only the calendar read waits.
- Many accounts are read at a steady pace. Each workspace has a fair share of the capacity Hive’s providers share between workspaces, and half of it is kept for your AI clients. A pass with more accounts than that share allows a minute reads them a few seconds apart instead of leaving the rest for next hour, carrying on where it stopped. A read that still cannot get its turn in time waits for the next pass.
- Very large workspaces take turns. A pass that runs out of time picks up next hour with the members it did not reach, so nobody is left out of the pass twice in a row.
- One account that cannot be read does not stop the others. If Hive cannot read one account on a pass (its provider is down, it is disconnected, or it lacks a permission), Hive skips that account and reads every other account, yours and other members’.
- A provider outage during the account check is not a disconnect. Before it reads an account, Hive confirms with the provider which account the connection signs in as. If the provider or Hive’s connection service does not answer that check (an outage, a rate limit or a timeout), Hive tries again on the next pass. It does not treat the account as a different or broken one, and does not ask you to reconnect.
When background reads are off for one account
For SharePoint Online, Microsoft Teams and Intercom, Hive first checks each account before it reads it in the background. The check confirms the provider, that the account was connected in this workspace, and that the account grants the permission the read needs. If the check fails, the account keeps working in Ask and for actions, but Hive does not read it in the background. Only the account’s owner sees why, in the tile’s Accounts panel:How Hive notices a broken connection
Hive marks an account Disconnected on its own only when it has confirmed the connection itself no longer works: the provider reports the grant as revoked or expired, the provider refuses to refresh the account’s sign-in, or a fresh check of the account (a new sign-in refresh, then the provider’s own “who am I” read) is refused. A sign-in refresh that fails only because the provider could not be reached (a dropped network connection, a timeout, or a server error or rate limit on the provider’s side) never disconnects an account. Hive and its connection service try the refresh again later, and the account keeps working once one succeeds. Any other refusal to refresh is treated as a sign-in that no longer works. A provider refusing one read (a 401 or 403 on one page or record) is not enough. That often means the account lacks the permission or scope for that particular data, so Hive tells you which data it could not read, that it may be missing a permission or need reconnecting, and keeps reading everything else from the account. For an app Hive has no account check for, signed in with a key or token that never refreshes (an API key, a personal access token, or an OAuth token that never expires), the connection is marked Disconnected when the app refuses (with a sign-in error) two different account-wide reads Hive makes on its own, or the same read three times with no successful read in between, within ten minutes. Several reads refused at the same moment count once for a token that never expires. An app whose OAuth sign-in expires and refreshes is judged by a fresh sign-in refresh alone, so refusals of reads already in flight never disconnect it. Only reads that name no particular record, workspace or board count, and a refusal the app says is about a missing scope or permission never counts. An Ask read counts only when it is one of the app’s verified reads; a read Ask had to guess never counts. Rate limits, provider outages and timeouts never disconnect an account. A disconnected account stops being read until its owner reconnects it. Hive never reconnects an account on its own. When the owner reconnects it, the account comes back as soon as the sign-in finishes, even if the browser tab is closed before Hive confirms it. Hive brings it back only when it is the same account that was connected before: signing in as a different account leaves the old one disconnected. Reconnecting the same account keeps it in every workflow and agent that uses it, so you do not need to choose it again or redeploy a workflow. A workflow run that failed while the account was disconnected is not run again: an app event that arrived in that time does not start its workflow a second time, so deal with those items in the app itself. You can also reconnect from Ask. When an answer could not read a tool because your own account for it needs reconnecting, it can offer Reconnect with the tool’s name and, when Hive knows it, the account’s, such as Reconnect Gmail (you@company.com). Each broken account gets its own button, and each repairs that same account in the same way, so its workflows and agents keep working. This is rolling out gradually; see Connect or reconnect an app from the answer.A provider the workspace has not enabled
Hive reads an account in the background only while its provider is enabled for the workspace. An Owner or Admin enables a provider by connecting it (see Accounts and autonomy). If an account’s provider is not enabled, Hive skips that account on each background sync, including document and mail sync, and keeps reading every other connected account, yours and other members’. The skipped account is not marked Disconnected and its owner gets no connection alert. It is read again on the next sync after an Owner or Admin enables the provider. Ask does not offer to connect or reconnect an app whose provider is not enabled: ask an Owner or Admin to enable it.A temporary permission or not-found error on one SharePoint library or OneDrive does not remove
that library’s documents from Hive. Hive keeps them, retries the library on the next sync and
reports the sync as partial. Documents are removed only when a complete scan of the account’s
sites no longer lists the library.
Connection alerts
When an account stops working, its owner gets one notification in the bell, for example Outlook disconnected: “Hive can’t read your Outlook data until you reconnect it. Open Integrations and select Reconnect.” Its Reconnect link opens Integrations.- It is sent once per incident, not on every failed sync.
- It names the provider only, never the account address or the provider’s error.
- An account with no individual owner notifies the workspace owner Inbox instead.
- Mute these alerts in your own bell with Connection problems under Settings → Notifications.
- the workspace has no Slack install, or installed Slack before these scopes existed (reinstall the Slack app to enable it — see Slack and Teams);
- the member’s Slack address differs from their Hive sign-in email;
- the account has no individual owner.
If connecting fails
When the provider’s sign-in window does not finish, Hive says why when it can, and what to do next. The same message shows in setup’s Connect your work step, under the app it is about, and as a notice on Integrations. Hive never shows the provider’s own error text.
In setup, a connection that was cancelled or failed has its own Retry button, and any app
that did connect stays connected. See
Connect your work.
Fix a connection
1
Reconnect
Select Reconnect on the tile and sign in to the provider as the same account. Hive repairs
the account in place: the tile, its Read, Draft or Act level, its position in the account
order and its history stay as they are, and no second account is added. If several accounts
of one provider need reconnecting, Reconnect opens the Accounts panel, where each broken
account has its own Reconnect button.
2
Check access in the provider
Make sure the account still exists, still has access to the data, and that an administrator
on the provider side has not revoked Hive’s app.
3
Check provider-specific settings
For PostHog, the region must match your project. For Sage 200, the account needs API access
and Company Access. See Connect an app.
What you can still search after a reconnect
Reconnecting a mailbox as the same account keeps the mail Hive had already read from it searchable in Ask and Knowledge, including after you disconnected it and connected it again. Hive moves that mail to the new connection once it has confirmed the new connection is the same provider account and has the same owner, so it stays searchable while the new connection reads the account again. Anything it reads again is not shown twice. It stays private to you: no other member can find it. For a file store (OneDrive, SharePoint, Google Drive, Dropbox), files come back into search as the reconnected account reads them again, so a file deleted or shared away from you while the account was disconnected never comes back. Google Drive and Dropbox files Hive already has are recognised rather than downloaded again. When the new connection finishes its first full read, it tidies up as a freshly connected account would: an item that is no longer in the account, or no longer in what Hive reads from it (for Gmail, mail from the last 365 days outside spam and trash), is removed from search. This applies only when Hive knows which account both connections signed in as. If you sign in as a different account, or Hive never recorded which account the old connection used, what the old connection read stays out of search, and the new connection reads the account from scratch. If a signal source is having trouble, its tile here and its read status on the Signals Matters tab show it, so an empty feed is never mistaken for an all-clear. See Matters.MCP server and webhook status
- MCP servers show Live when connected, Authorizing while OAuth is pending, or an error. Resync re-discovers the server’s tools. See MCP servers.
- Incoming webhooks show Awaiting events until the first signed event arrives, then Receiving. See Incoming webhooks.
- Slack and Teams channels show as active, needing re-authorisation, or disconnected. See Slack and Teams.
Disconnect a tool
Open the tile and choose to disconnect. Hive confirms first: “Hive will stop reading from it. Nothing you have already approved is undone.” When a provider has several accounts, the prompt says all of them are removed and Hive stops reading from every one. When you disconnect a managed connector:- Hive deletes the connection and stops reading from the account.
- For a file store or mailbox, Hive deletes its sync position for the account — how far it had read the files or mail, and the items it still had to retry — so nothing resumes from it. If you connect the account again, Hive starts with a full scan.
- Mail and files Hive had already read from the account stop appearing in search straight away. Google Drive and Dropbox files stay searchable while another account of the same app is still connected in the workspace, for the people who can open them. If you connect the same account again, the rest come back (see What you can still search after a reconnect).
- Every open signal the account was feeding is withdrawn, including a cross-tool signal that drew on it. Their pending approvals are removed, so nothing can act through an account that is gone, and their notifications are withdrawn with them. A withdrawn signal reads Source disconnected instead of Dismissed and does not count as your team dismissing it. A signal whose action is already running is not withdrawn. See When Hive loses access to a source.
- Work Hive had queued for those signals — a plan, an automatic action or a retry — stops before it sends anything. The same happens if you re-authorise an account while work is queued: it stops rather than run under access that changed after it was queued.
- Actions that already ran are not undone. Undo reversible ones from the audit log if you need to.
- Each account is revoked separately. If one fails, the others stay connected and you can retry.
- MCP server — its pinned tools and all stored secrets are deleted, and the connection is marked Disconnected.
- Incoming webhook — delete it; the URL stops accepting events.
- Channel — the installation is revoked. For Teams, the app and consent on the Microsoft side remain until you remove them there.
Related
Troubleshooting
Fixes for common problems.
Accounts and autonomy
Ownership and levels.