> ## Documentation Index
> Fetch the complete documentation index at: https://docs.cortex-labs.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Run a workspace app

> Open the intended app instance, follow its configured actions and verify outputs and decisions.

This exercise runs one available app against your own small permitted input. Choose an app whose purpose and expected result you understand.

You need workspace access, app execution permission and access to its displayed data. Starting an app does not supply missing data or review authority.

<Steps>
  <Step title="Open the workspace instance">
    Select the app in the intended workspace. Check its name, purpose and local configuration.

    An app template is reusable. This instance owns its workspace settings and runs. Confirm that you opened the intended instance.
  </Step>

  <Step title="Check the input source">
    Inspect selected datasets, files and parameters where your access permits it. Ask an instance updater to repair missing or incompatible settings.

    Instance parameters configure a run. A form answering a human wait changes that wait. They are different inputs.
  </Step>

  <Step title="Start one run">
    Use the app's run control once. Retain the run identifier and inspect its instance, template and workflow versions.

    Follow the configured tabs or steps. A completed screen element does not by itself establish successful publication.
  </Step>

  <Step title="Use the supported elements">
    Open tables, evidence or documents as the app presents them. Follow source links before interpreting a summary.

    For a model-step chat, inspect its transcript. Use send controls only where the configured element supports an interactive workspace-agent chat.

    If a wait is assigned to you, inspect its subject and settle it through the offered decision or answer controls. Read-only mode can disable those actions.
  </Step>

  <Step title="Inspect the outcome">
    Check failed or skipped blocks, partial data, open waits and final outputs. Inspect publication separately from drafting and run completion.

    Record remaining limitations. Reopening the existing run view does not start another inference run.
  </Step>
</Steps>

<Frame caption="Illustrative app assembled from document, table and log elements. Inputs and controls depend on the app’s template.">
  <img src="https://mintcdn.com/cortex-ai/vcVF551foITpk__t/images/docs-20261019-11-workspace-app.png?fit=max&auto=format&n=vcVF551foITpk__t&q=85&s=b93a577b573afca7228dfd0ed3f9e3d0" alt="Run a workspace app mockup." width="1440" height="900" data-path="images/docs-20261019-11-workspace-app.png" />
</Frame>

## When the app cannot start

Check whether the instance is archived, unpublished, read-only or incompatible with its template. A locked workspace or archived workflow can also block execution.

A template update affects new runs when the instance follows latest. Past runs retain their versions. Ask the instance owner to inspect a new incompatibility before retrying.

Use [Understand run outcomes and costs](/vantage/run-apps-and-workflows/run-outcomes-costs) for progress and retries. Use [Decide an assigned review](/vantage/review-and-publish/decide-review) for output-review decisions.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.