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

# Configuring Agents

> Set agents at the organisation level, override per workspace, and roll out with confidence.

## Two Levels of Configuration

Agent configuration follows a simple model. You set organisation-level defaults that apply everywhere, then optionally override them for specific workspaces. This gives you central control with the flexibility to treat some teams differently.

## Organisation Defaults

Your organisation settings define the baseline for each agent: its role, and what it acts on. For Gatekeeper that is your default rule and exceptions. For Data Protector that is the categories you enable and any instructions you assign. These defaults apply to every workspace unless a workspace overrides them.

Set your defaults to reflect your most common needs. Workspaces that need different treatment can be configured on their own.

## Workspace Overrides

Each workspace handles an agent in one of three ways.

| Mode             | Behaviour                                                                    |
| ---------------- | ---------------------------------------------------------------------------- |
| **Inherited**    | Uses the organisation default exactly. This is the case for most workspaces. |
| **Configurable** | Sets its own role and settings for the agent, independent of the default.    |
| **Disabled**     | Turns the agent off for this workspace. No review happens.                   |

<AccordionGroup>
  <Accordion title="Use Inherited for most workspaces" icon="building">
    Keeps your configuration central and easy to manage.
  </Accordion>

  <Accordion title="Use Configurable for special cases" icon="sliders-horizontal">
    A workspace handling sensitive customer data might run Data Protector as an Enforcer while the rest of the organisation observes.
  </Accordion>

  <Accordion title="Use Disabled when an agent does not apply" icon="circle-off">
    Turn an agent off for a workspace where its focus is genuinely irrelevant.
  </Accordion>
</AccordionGroup>

## Rolling Out

A phased rollout reduces disruption and builds confidence.

<Steps>
  <Step title="Start in Observer" icon="eye">
    Run both agents as Observers. You get full visibility into what they catch, with no impact on your team. Stay here for a week or two.
  </Step>

  <Step title="Review and tune" icon="list-checks">
    Look through [Assessments](/agents/assessments) to understand your usage. Set your Gatekeeper policy, enable the Data Protector categories that matter, and add [instructions](/agents/instructions) for the specific cases.
  </Step>

  <Step title="Promote to Enforcer" icon="shield-check">
    Move agents to Enforcer for the workspaces where the risk justifies it. Begin with your highest-risk workspaces and expand as confidence grows.
  </Step>

  <Step title="Override where needed" icon="layout-grid">
    Once your defaults are stable, set workspace overrides. Stricter enforcement for high-risk teams, lighter touch elsewhere.
  </Step>
</Steps>

## Setup Checklist

| Step                          | Description                                                        |
| ----------------------------- | ------------------------------------------------------------------ |
| Set organisation defaults     | Choose a role and settings for each agent.                         |
| Identify high-risk workspaces | Find the workspaces handling sensitive or regulated data.          |
| Configure overrides           | Use Configurable mode where a workspace needs different treatment. |
| Disable where irrelevant      | Turn off an agent for workspaces it does not apply to.             |
| Monitor and adjust            | Review assessments regularly and refine over time.                 |

***

<CardGroup cols={2}>
  <Card title="Understanding agents" icon="shield-check" href="/agents/understanding-agents">
    Roles, outcomes, and how verdicts are decided.
  </Card>

  <Card title="Instructions" icon="fingerprint" href="/agents/instructions">
    Add your own rules for specific scenarios.
  </Card>
</CardGroup>