Troubleshooting

Most of what goes wrong in Tempio is a platform saying no, and most of it is readable off a connection's chip and the line under its name. This page decodes both.

The six states a connection can be in

Every connection shows one chip with its state: green for Connected, amber for Syncing, red for Needs attention, and grey for Not synced, Paused and Disconnected. Most states add a line under the name with the detail. Connections has the same table with the lines a Reddit, X or Microsoft Advertising card shows.

Chip	Line under it	What to do
Not synced	None. Connected; no sync has started.	Nothing. The schedule picks it up on its own: connections on a paid plan sync hourly, free workspaces daily. Sync now starts it immediately.
Syncing	`3 min`, then the stage it is in, with counts that move as rows land. A request whose run has not opened yet reads `Queued, starting shortly`.	Wait. A first sync of a large account with 90 days asked for takes 10–20 minutes.
Connected	`Synced 12 min ago`. The last run finished cleanly.	Nothing. A restated badge beside the line means that run rewrote a lot of already-stored history, which is the platform correcting itself.
Needs attention	`Last sync finished with errors`, then `2/3 accounts synced`. Some accounts landed and some did not, so the numbers are fresh but incomplete.	Sync now. If it repeats, the account that failed has probably lost access on the platform's side.
Needs attention	A failure headline, then the platform's own words. The run failed. The headline says what it was and the line under it is what the platform said.	Read the table below.
Paused, or Disconnected	`Syncing is off` under Paused; nothing under Disconnected. Switched off, on purpose. Nothing new lands, and no draft applies to it.	Stored history stays. Set the connection active again to resume.

What a failure headline means

Headline	What happened	What fixes it
This connection needs to be reconnected.	The platform no longer accepts the credentials Tempio holds — an expired token, or access revoked on the platform's side.	Reconnect on the row. Nothing already stored is deleted or re-read from scratch.
The platform refused access to this account.	The grant is alive but does not carry the access this account needs.	Check the permissions granted to Tempio on the platform, then reconnect.
The platform asked us to slow down.	A rate limit.	Nothing. Syncing resumes on its own.
The platform rejected the request.	The platform refused a specific call as invalid.	Read the line under it — it is the platform's own message, naming its own field.
The platform had a problem on its side.	A platform-side fault.	Nothing. Tempio retries.
Something we were tracking no longer exists on the platform.	An entity Tempio holds was deleted on the platform.	Nothing. The next sync settles it.
Tempio can't do that on this platform yet.	The operation is not one this platform's adapter implements.	Platforms says what each one can express.
We couldn't reach the platform.	A network failure.	Nothing. Tempio retries.
We couldn't confirm whether the last change landed.	A call went out and the answer never came back.	Nothing yet. It is verified before anything is retried.
The last sync didn't finish.	The fallback, when the failure does not match any of the above.	Sync now; if it repeats with the same line, send it to support.

Reconnect re-runs the platform's sign-in and keeps every row Tempio already holds. It also revives the connections the failure parked — but a connection you paused or disconnected on purpose stays as it is, because reconnecting a grant should never quietly switch syncing back on.

A grant in error

A connection fails because its grant failed, and grants fail for a handful of reasons the platform tells us: the token expired, the token was revoked, a permission was revoked, the ad account was closed, the platform had an error, or somebody asked the platform to delete or deauthorise the app. The first several all end in the same place — the grant has to be re-authorised, which is what Reconnect does. An account closed on the platform is the one that does not: nothing on this side reopens it.

One grant can carry several connections. When the grant goes, they all stop together and they all come back together.

The numbers look wrong

A rate shows —. That is not zero. A ratio with a zero denominator has no value — no conversions yet means no cost per conversion — and Tempio renders that as a dash rather than as a number that would read like good news. Metrics has the rule.

Reach is empty. Only some platforms report reach on the daily grain. Empty means the platform did not report it; zero would mean it reported that nobody was reached.

Yesterday changed overnight. Expected. Every sync re-reads a trailing window — 28 days on Meta, 30 on Google, 14 on TikTok — so late conversions and platform restatements land as corrections. A large rewrite shows as restated on the connection.

Today looks low. Today is still being counted, on the platform's side and on Tempio's. Compare complete days.

A total is called an estimate. The scope covered more than one currency, so the figures were converted at published rates and labelled, with the date of the oldest rate used. Single-currency scopes are never converted and never labelled.

The dashboard is empty and the connection says synced. Check the range and channel chips at the top of the dashboard; both persist as you move around.

A draft that will not go

What you see	What it is
A guardrail sentence with two amounts in it	The change exceeded a cap. The sentence names the amount and the cap, both in the account's currency. Guardrails
`Pause (draft pending)`, and the menu item disabled; or, on a campaign's or an ad group's page, a disabled button whose tooltip says a draft is waiting for approval	A change request for that row and that operation is already at the gate. Approve or reject it, and the row opens again.
A batch that stopped partway	A row failed or half-landed and the batch stopped there. What landed stays landed; the rows behind it went back to waiting. Edit the failed row and approve the group again — it resumes from where it stopped. Drafts
Still `applying` after 15 minutes	The apply stopped answering and the outcome is genuinely unknown. Check the platform yourself, then close the row with Mark failed, which records that the outcome was never confirmed.
A refusal quoting the platform's own field names	The platform's dry run rejected the payload before anything was sent. Fix the payload on the draft and it goes back to waiting.
Nothing applies on this connection at all	The connection is not active. That is the kill switch, and it blocks every apply written against it whatever cleared them.

Limits

Message	What it means
You've used every workspace your plan includes	The workspace allowance. A different name will not help.
Connecting is refused before the platform's sign-in opens	The ad-connection allowance for the plan.
A new key cannot be minted	The MCP key allowance. Revoked keys do not count toward it, so revoking an unused one frees the slot.
An invitation is refused	Seats, counting pending invitations.
You've reached your agent message limit	Either the monthly pool is spent or the daily brake bit. The usage card says which.

Billing has all six allowances.

If none of that is it

Two things make a report quick to act on: the connection or draft's URL from your address bar, and the exact line you were shown, copied rather than described. The line under a failure headline is the platform's own words, and it is usually the sentence that identifies the problem. Write to hello@tempio.ai.