---
title: "Start a QA run and stop it | Proofline Help"
description: "Start a run from a setup's page when you want the agent to explore the setup now instead of at its next scheduled time."
url: https://proofline.sh/help/qa-start-a-run/
---

[Help center](/help/)

# Start a QA run and stop it

Start a run from a setup's page when you want the agent to explore the setup now instead of at its next scheduled time. Starting and stopping runs needs **Operate Agentic QA**, which organization administrators hold and a [custom role](/help/roles/) can grant.

## Start a run

1.  Open **Agentic QA** and select the setup.
2.  In the setup's **Runs** section, select **Start run**.

The button reads **Starting** while Proofline admits the run, and then the new run's page opens. The run follows the setup as it is saved at that moment, so save any edit first. A later edit does not change the run's objective, journeys or limits, though a change to its sites or forms ends it.

A start holds the setup's **Allocation for one run** from your organization's credits, and the run spends from that hold. A setup runs one run at a time: if the setup already has a run that has not ended, **Start run** opens that run instead of starting another. Selecting the button twice also opens a single run.

## When Start run is unavailable

The **Runs** section shows **Start run** but does not let you select it, and gives the reason beneath it, when no run of the setup can start:

-   **Agentic QA runs are not enabled yet.** Runs are not available for your organization. You can still prepare and edit the setup, and its schedule starts no run either.
-   **No browser provider is configured.** Proofline has no browser to run the setup in.
-   **This organization is read-only.** Nothing in the organization can start or stop a run.

An archived setup offers no **Start run**.

## When a start is refused

Proofline checks the start again when you select **Start run**. A refused start begins no run and holds no credits, and its reason appears in the **Runs** section:

-   The organization already has two runs in progress. Start this one when one of them ends.
-   The run's allocation is more than the organization's available credits. The reason gives both amounts; add credits, or lower the setup's **Allocation for one run**.
-   The organization's spend limit is reached.
-   An earlier run's cost is still being reconciled. No run starts until that finishes.
-   Your role lacks **Operate Agentic QA**. The reason names the capability; ask an organization owner for it.

A setup with an environment also needs that environment ready to take the run. When it is not, the reason says what is missing. A link beside most of these reasons opens the setup's **Environment** section: **Follow the command** when an environment command is still in progress, **Start the environment** when it is not running, or **Review the environment** when its repository has not passed readiness or it cannot take the run's agent session.

## Follow a run

The run's page shows its state beside the title: **Queued**, **Running**, **Paused**, **Stop requested** or **Finishing** while it lasts, then **Ended**, **Stopped**, **Failed** or **Outcome unknown**. A run in an environment is **Paused** while it waits for that environment; [resume it after its environment stops](/help/qa-resume-interrupted-environment/).

The setup's **Runs** section lists its runs newest first. Each reads **Manual run by** the person who started it or **Scheduled run**, with its state, when it started and how long it ran. Select one to open its page.

No state says whether your site passed. Read the run's coverage, journeys and findings as [What a run reports](/help/agentic-qa/#what-a-run-reports) describes.

## Stop a run

Open the run and select **Stop**. The button is offered while the run is **Queued**, **Running** or **Paused**, and reads **Stopping** while the request is sent. The run then reads **Stop requested** until the agent ends before its next step, and **Finishing** while it settles its credits. Once it has ended, it reads **Stopped**, and **How it ended** says **The run was stopped**.

A stopped run keeps its page and what it recorded before the stop, and **Credits** shows what it used; [explain the credits shown on a QA run](/help/qa-understand-run-charges/) when part of them is held. Confirmation starts nothing after a stop, so a problem the run had not confirmed stays unconfirmed and is never filed.

Stopping a run leaves the setup's schedule as it is. To stop future scheduled runs, [pause the schedule](/help/qa-pause-a-schedule/).

## Related guides

-   [How Agentic QA works](/help/agentic-qa/)
-   [Pause scheduled QA runs without losing the schedule](/help/qa-pause-a-schedule/)
-   [Explain the credits shown on a QA run](/help/qa-understand-run-charges/)
