A tool failed to run

Summary#

A tool run passes through a fixed sequence of checks before any data is fetched, and the credit is taken at one specific point in that sequence. Everything that fails before that point is free. Everything that fails after it has already cost you the credit. This article walks the whole sequence, gives the message you see at each stage, and states plainly what happened to your balance.

Symptoms#

The result area under the tool shows one of these instead of a report:

• A red block with a message, or the generic Something went wrong.
• A neutral block reading Monthly limit reached — upgrade to continue.
• A red block reading Please sign in to run this tool.
• The loading line Analyzing… this can take a few seconds for live data. that never resolves.
• A message naming the AI provider, for example Too many requests right now - please wait 10 seconds and try again.

Diagnosis#

The run sequence, stage by stage#

Read down the table until you find your message. The Credit column is the answer to the only question that usually matters.

Stage 1 to 3: nothing has been charged#

Please sign in to run this tool. Your browser has no session. Reload, and if you are sent to the sign-in page, sign in again. See The dashboard is stuck loading.

The You already ran this panel. You have run this exact query with these exact settings before, and the saved copy is still available. The panel offers Open saved result, Run fresh (N credits) and Cancel. Open saved result costs nothing, Run fresh charges the current price of the tool. The panel also tells you when the earlier run happened and what it cost then, which is not always the same as what it costs now.

This tool is temporarily unavailable. Please try again shortly. That individual tool has been switched off, usually while a provider problem is being worked on. Other tools are unaffected. Try again later, and check 503 and service-unavailable errors if it persists for more than an hour.

Stage 4: the recommendation allowance#

The Get Recommendations button on a result has its own monthly allowance, separate from your credits, and its own minimum plan. If you are below that plan, the panel shows Recommendations are a paid feature with a See plans button. If you are out of allowance, it shows You've used your recommendations with an Upgrade for more button.

If it passes, the recommendation allowance is spent before the AI model is called and is not returned if that call then fails. The same click also costs one ordinary credit. See Get Recommendations.

Stage 5: identity, plan and credits#

These all return before anything is charged.

The last two are deliberate. When we cannot confirm what you are entitled to, we stop rather than guess, so nothing is spent on your behalf without being counted.

Stage 7: the input was rejected after the charge#

For the technical tools these checks cost nothing, because those tools are priced at zero credits. Two tools in that group are priced at one credit, so a rejected address there has already cost you that credit.

Stage 8 and 9: the provider or the model failed#

By this point the credit is spent. These messages are rewritten into plain language before you see them:

Growth Actions has its own messages#

Growth Actions reads your existing saved results rather than fetching new data.

• Sign in to generate Growth Actions. means no identity reached the server.
• If it finds no saved results to work from, it returns a "what to run first" state and charges nothing.
• Growth Actions storage is unavailable right now. is a temporary storage problem.
• Could not generate the queue with a reason underneath means the model call failed after the credit was taken.

The run never finishes at all#

If Analyzing… this can take a few seconds for live data. stays on screen indefinitely, the response never came back. Heavy reports legitimately take a minute or more; a full article draft is expected to take around 60 to 90 seconds. Beyond a few minutes, treat it as failed: reload the page and check whether the credit was counted (see below) before running it again.

Resolution#

1. Read the message and find it in the tables above. The message tells you which stage stopped, and therefore whether you were charged.
2. Retry once for anything marked retryable. Rate limits, region flags and unreadable model answers usually clear on a second attempt.
3. Check whether the credit was taken. The credit pill in the top bar reads <used> / <quota> reports. It refreshes after every successful premium run and on page load, so reload before comparing. Account → Usage shows the same figure with a per-tool breakdown. See Tracking your usage.
4. Do not re-run repeatedly. Each attempt that gets past the gates is charged again, even if it fails immediately afterwards.
5. Open the saved copy instead where one exists. If the You already ran this panel offered a saved result, opening it is free and gives you the earlier report. Saved results are kept for 90 days. See Saved Work.
6. Ask for the credits back if a provider failure cost you several. We can reset the month's usage or credit an amount back to your counter. Include the tool name, the query, the approximate time, and the exact message. See Contacting support.
7. If a whole category of tools fails, check 503 and service-unavailable errors before assuming it is your account.

Prevention#

• Use the saved copy when the duplicate panel offers one. It is free and instant, and for slow-moving data it is as good as a fresh run.
• Check your address before running an address-based tool. A typo that gets past the gates is charged like any other run.
• Keep an eye on the credit pill before starting a batch of heavy reports. Running out mid-batch leaves you with partial work.
• Space out light-tool runs. The hourly fair-use limit is 100 light calls per clock hour. Normal use never reaches it; scripted or repeated clicking does.
• Run the tools that Growth Actions depends on first, so it has something to synthesise and does not return the "nothing to work from" state.
• Report repeat failures rather than retrying. A tool that fails twice for the same input will usually fail a third time, and each attempt past the gates is billed.