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

# Adding Agents from the Marketplace

> Review a ready-made Agent, add it to your workspace, and connect the services it needs.

The Marketplace lets you start with an Agent built for a specific outcome instead of configuring one from scratch. Open the listing, review how it works and what it can access, then add it and complete any guided setup.

<Note>
  Adding an Agent requires **Builder access** (workspace admin). If you do not have it, Erstan explains which permission is required. Team installation follows the permissions shown on the Team listing.
</Note>

## Review the Agent

From the [Marketplace](/marketplace/overview), open an Agent card. The header shows its outcome, installed state, and a truthful **Built by** byline. Erstan-authored Agents show **Built by Erstan**.

Use the four tabs to evaluate the Agent:

<Tabs>
  <Tab title="Overview">
    Read the intended outcome, who the Agent is for, and the most important usage and safety guidance.
  </Tab>

  <Tab title="Capabilities & access">
    Review the Agent's capabilities, human-readable tools and access, required connections, and declared read/write posture. Technical tool IDs and model settings are intentionally omitted.
  </Tab>

  <Tab title="Example requests">
    Start with practical requests curated for the Agent. These combine its built-in starters with related examples without exposing separate Prompt products.
  </Tab>

  <Tab title="How it works">
    Inspect the Agent's workflow and operating mode in plain language.
  </Tab>
</Tabs>

## Understand access and approvals

The detail page only makes a trust claim when the Agent declares enough information:

| Posture                           | What it means                                                                                                                |
| --------------------------------- | ---------------------------------------------------------------------------------------------------------------------------- |
| **Read-only**                     | The declared tools read data without making changes.                                                                         |
| **Writes require approval**       | The Agent can propose a write, but the action pauses for your approval.                                                      |
| **Workspace tool policies apply** | Access is controlled by workspace policy; the listing does not declare enough to infer read-only or approval-gated behavior. |

<Warning>
  Never treat missing tool metadata as proof that an Agent is read-only. Your NetSuite role, per-user authorization, and workspace approval policy remain authoritative. See [Security & write safety](/security/overview).
</Warning>

## Add the Agent

The primary action reflects the Agent's operating style:

* **Add Agent** creates an editable conversational Agent in your workspace.
* **Create Automation** opens the visual builder with the automation graph prefilled.
* **Install solution** installs a single-Agent solution together with its curated capabilities, requests, and templates.

Existing installed listings show the appropriate open, update, or installed state instead of creating a duplicate.

## Complete guided setup

When setup is required, Erstan walks through the Agent's introduction, instructions, required connections, and completion state.

<Steps>
  <Step title="Review the introduction">
    Confirm what was added and how the Agent is intended to be used.
  </Step>

  <Step title="Read the operating guidance">
    Review any Agent-specific instructions before continuing.
  </Step>

  <Step title="Connect required services">
    Each required connection reports **Connected**, **Not connected**, or **Checking** and provides the relevant connection action.
  </Step>

  <Step title="Start using the Agent">
    Finish setup and open the installed Agent.
  </Step>
</Steps>

Each member still authorizes their own NetSuite user, so the Agent works with that person's role and permissions. See [Authorizing your NetSuite user](/connectors/netsuite-authorize).

## After installation

From the installed Agent you can:

* Try one of its example requests or start from chat.
* Run it on demand or configure the supported trigger.
* Inspect [run history and traces](/agents/run-history-traces).
* Review any approval-gated action before it executes.

<Check>
  If sign-in is required during installation, Erstan returns you to the listing afterward so you can resume.
</Check>

## Next steps

<CardGroup cols={2}>
  <Card title="Marketplace overview" icon="store" href="/marketplace/overview">
    Browse the combined Agent and Team catalog.
  </Card>

  <Card title="Installing Teams" icon="people-group" href="/marketplace/add-skills-bundles">
    Review members, capabilities, example requests, and safety before installing a Team.
  </Card>

  <Card title="Approving AI actions" icon="shield-check" href="/chat/approvals">
    See how write actions pause for review.
  </Card>

  <Card title="Running Agents" icon="play" href="/agents/run-and-tasks">
    Run installed Agents and turn their output into work.
  </Card>
</CardGroup>