# Get started (/docs/getting-started)



This guide takes a new workspace through one complete customer workflow. You need a product endpoint that Invokeable can reach and a nonproduction account that assessment runs may use.

<Callout type="warning" title="Use a sandbox or staging environment">
  An assessment can create, update, send, or delete records when the journey
  requires it. Keep live customer data out of the connected test environment.
</Callout>

## Before you begin [#before-you-begin]

* Use an Admin account for initial setup.
* Prepare a GitHub repository, MCP endpoint, OpenAPI URL, or exported contract file.
* Know whether the target environment is a sandbox, staging system, or production system.
* Have any required token or API key available. Invokeable seals it in your browser before it is sent.

<Steps>
  <Step>
    ### Create or open your workspace [#create-or-open-your-workspace]

    Sign in and create a workspace if your organization does not have one. A workspace contains your product connections, agents, journeys, runs, findings, and members.
  </Step>

  <Step>
    ### Connect your product [#connect-your-product]

    Open **Catalog** and choose **Connect product**. Connect the GitHub repository that contains `invokeable.yaml`, add an MCP or OpenAPI URL, or import a contract file.

    <DocImage src="/images/product/connect-product.jpg" alt="Connect your product form with GitHub, URL, and file choices" caption="Use a sandbox or staging connection for product actions that can make changes." />
  </Step>

  <Step>
    ### Supply a credential if requested [#supply-a-credential-if-requested]

    Paste the token or API key into the secure credential step. The browser encrypts the credential for the active workspace, request, and connector before transmission. If a connection needs no credential, Invokeable skips this step.
  </Step>

  <Step>
    ### Review the discovered actions [#review-the-discovered-actions]

    Check each action's purpose, access level, risk, confirmation requirement, retry behavior, and verification method. An action marked **Needs setup** cannot be used safely until its missing information is resolved.
  </Step>

  <Step>
    ### Check connected agents [#check-connected-agents]

    Open **Agents**. At least one agent must be active and compatible with the capabilities required by the journey. Open an agent to review limitations when its status needs attention.
  </Step>

  <Step>
    ### Create a journey [#create-a-journey]

    Open **Journeys**, choose **Create journey**, and describe one customer task. Include the intended result and any behavior that must not happen.

    ```text title="Example journey description"
    A customer asks to change from the Starter plan to the Growth plan.
    Confirm the new plan and billing date. Do not create a second subscription.
    ```
  </Step>

  <Step>
    ### Review the test plan [#review-the-test-plan]

    Confirm the success conditions, protected behavior, failure response, and coverage. Edit anything that does not match the customer task before you run it.
  </Step>

  <Step>
    ### Run the assessment [#run-the-assessment]

    Select **Run test**, review the test count and safety limits, then select **Start test**. You can leave the page while Invokeable runs and verifies the assessment.
  </Step>

  <Step>
    ### Review the result [#review-the-result]

    Open **Findings**, then select **Test results**. The summary shows each check as passed, failed, or inconclusive. Use **Activity** when you need the sequence of actions and observations.
  </Step>
</Steps>

## What completion means [#what-completion-means]

A completed run is not automatically a successful run. Invokeable separates execution state from the verified outcome:

* **Passed** means the expected state and protected behavior were verified.
* **Needs attention** means one or more checks failed or a repeatable product problem was found.
* **Couldn't complete** means the assessment stopped, failed, or lacked enough evidence for a trustworthy decision.

The agent's final message is context. The verified product state decides the outcome.

## Next steps [#next-steps]

* [Connect your product](/docs/guides/connect-product)
* [Create a journey](/docs/guides/create-journey)
* [Review test results](/docs/guides/review-results)
* [Complete the same workflow with the CLI](/docs/cli/first-assessment)