# Automation runs and troubleshooting > See every automation run, find where a run stopped, retry a failed step safely, ask Orbit for a repair and control automation spending. Canonical: https://help.spacebrain.ai/automations/runs-and-troubleshooting/ Last modified: 2026-09-25T00:00:00.000Z Each time an automation starts, Spacebrain records a run: who it ran for, which steps ran and what happened. This guide shows you how to check runs, understand a failure and fix it without sending the same message twice. It's for anyone who looks after live automations. ## Before you begin [#before-you-begin] * **Permission:** anyone who can open Automations can view runs. Retrying a step or preparing a repair needs editing permission. * **Know the record:** it helps to know the lead, person or company you're checking, and roughly when the event happened. ## Where to find runs [#where-to-find-runs] * **Automations → Runs** lists runs from every automation in one place. * In **Automations → All Automations**, the **Runs** column shows each automation's run count. Select the count to see a run summary, then **View executions**. ## Read the Runs list [#read-the-runs-list] **Runs** shows "Monitor automation and email executions from one operational history." You can search, and filter by engine (**Automations** or **Email**) and by status. | Column | What it tells you | | ------------------- | ------------------------------------------------------------------- | | **Automation** | Which automation ran. | | **Engine** | **Automation** or **Email**. | | **Subject** | The type of record it ran for, such as a lead. | | **Status** | Where the run is now (see below). | | **Started** | When it began. | | **Finished / next** | When it ended, or when the next step is due. | | **Issue** | For a failed run, where it stopped, such as "Failed at Send Email". | ### Run statuses [#run-statuses] | Status | Meaning | | ------------- | -------------------------------------------------------------------- | | **Pending** | Queued and about to start. | | **Running** | Working through its steps now. | | **Waiting** | Paused on a **Delay** or waiting for something before the next step. | | **Paused** | On hold and not currently moving to the next step. | | **Completed** | Finished its path. | | **Failed** | Stopped because a step couldn't finish. | | **Cancelled** | Stopped before finishing and won't continue. | ## Open a run's details [#open-a-runs-details] Select a run to open **Execution Details**. You'll see: * The automation, **Trigger Event**, **Record Type** and **Record ID**. * **Started At**, **Completed At** and **Duration**. * **Execution Steps:** each step in the order it ran. Condition steps show whether they went down the **TRUE** or **FALSE** branch. Select **Show details** to see the step's **Input** and **Output**. * **Execution Logs:** messages recorded while the run happened. They update live while a run is in progress. * **Execution Context:** the data the run started with. ## Fix a failed run [#fix-a-failed-run] A failed run shows a **Failure recovery** section. Before you do anything, check what already happened. 1. **Find the first failed step.** Read its error message and its **Input**. 2. **Check what already happened.** Steps before the failure did run. If an email, text or call step ran, check the inbox or CRM activity before sending again. 3. **Fix the cause.** Common causes are a disconnected account, a missing email address or phone number, or a setting left blank. Select **Inspect automation node** to open the failed step in the editor. 4. **Choose how to recover:** * **Retry failed step:** continues from the failed step, using the same automation version. Steps that already succeeded are not repeated. * **Prepare AI repair:** opens **Repair an automation draft**. Orbit suggests a fix as a draft. It doesn't replay the failed run. Review the draft, run a dry test and publish when you're happy. > **Warning — Don't send a message twice** > > If the failed step sent a message or placed a call, Spacebrain may show **Check the communication outcome before sending again.** That means it can't prove a retry is safe. Check the inbox or call history first. Very old runs may show **This legacy run has no saved automation version, so it cannot be retried safely.** Fix the automation and let the next event run it instead. ## Worked example: a follow-up email failed [#worked-example-a-follow-up-email-failed] A run shows "Failed at Send Email" with an error about the sender account. 1. Open **Automations → Connections** and reconnect the email account. 2. Open the run. Check **Execution Steps**: **Add Tag** and **Create Task** completed, **Send Email** failed. 3. The email was never sent, so a retry is safe. Select **Retry failed step**. 4. The run continues from **Send Email**. The tag and task aren't created again. 5. Confirm the email appears in the lead's activity. ## Control spending [#control-spending] Some steps, such as AI calls and AI replies, use credits. You can set a monthly budget for each automation: 1. Open the automation in the editor and select its budget button. It shows the current state, for example "No budget" and the credits spent so far. 2. In **Automation admission budget**, set **Monthly budget · credits** and an **Email alert threshold · %**. 3. Turn on **Block new paid actions at the limit** if new runs should stop when the budget is used up. 4. Select **Save policy**. A budget stops new runs from starting. It doesn't stop runs that are already in progress. To see workspace-wide numbers, open **Automations → Usage**, which tracks active automations, configured nodes, executions and charged credits. ## If something goes wrong [#if-something-goes-wrong] | Symptom | What to check | | ---------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------- | | No run appeared | The automation is **Active**, the trigger matches the event, and any **Only this form** or **Condition** check lets the record through. Drafts never run. | | Run is stuck on **Waiting** | A **Delay** is counting down. **Finished / next** shows when the next step is due. | | Run failed at a messaging step | The sender account in **Automations → Connections**, the recipient's email or phone number, and consent or opt-out status. | | Run failed at a table step | The table still exists and its columns haven't changed. Open the step and select **Use reviewed schema** if asked. | | New runs stopped | The automation may be paused, or its budget may be blocking new runs. Check the budget button in the editor. | | **Automation runs could not be loaded.** | Select **Try again**. If it continues, check your connection or permissions. | For message delivery problems, see [Communications and automations troubleshooting](/troubleshooting/communications-and-automations/). ## FAQ [#faq] ### Does retrying repeat steps that already worked? [#does-retrying-repeat-steps-that-already-worked] No. **Retry failed step** continues from the failed step. Earlier successful steps are not repeated. ### Does pausing an automation stop runs already in progress? [#does-pausing-an-automation-stop-runs-already-in-progress] Pausing stops new runs from starting. Check **Runs** for anything still **Waiting** and decide whether it should continue. ### Can I test without creating a run? [#can-i-test-without-creating-a-run] Yes. The dry **Test** in the editor and classifier tests don't create real runs or send anything. See [Build an automation in the editor](/automations/workflows/). ## Next step [#next-step] Improve a draft with [Build an automation with Orbit](/automations/build-with-orbit/), or go back to the [Automations overview](/automations/).