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

# Agent Creation

> Create, configure, and approve organization-managed agents.

Create organization-managed agents with a clear purpose, least-privilege defaults, and an explicit approval path.

## Open the agent manager

1. Open **Organization → Agents**, or visit /my-org/agents.
2. Review the agent list, pending approvals, and current execution counts.
3. Select **Create Agent**. Use **Create with AI** for the assisted creation flow.

![Agent management](https://mintcdn.com/illuminarescientiainc/M60tJR457eGexcID/images/05-agent-management.png?fit=max\&auto=format\&n=M60tJR457eGexcID\&q=85\&s=86c6ec273a5f38cf165afa5b92aac62a)

## Complete the creation flow

1. On **Identity**, enter a clear **Agent name**, choose a **Role**, and select an avatar.
2. On **Purpose**, describe what the agent does and add optional system instructions.
3. On **Classification**, choose up to six categories, add comma-separated tags, and set visibility.
4. On **Review**, confirm the details and select **Create agent**.

![Agent creation: Identity](https://mintcdn.com/illuminarescientiainc/M60tJR457eGexcID/images/06-agent-creation.png?fit=max\&auto=format\&n=M60tJR457eGexcID\&q=85\&s=ae5977fea0acf31ab5ed7e5857aa424c)

## Configure and approve the agent

1. Open the saved agent’s details page.
2. Configure the provider and model before activation.
3. Configure tools, knowledge, guardrails, capabilities, and access.
4. Submit the agent for organization approval.
5. Activate it only after an authorized approver completes review.

New agents start with no optional capabilities, empty tool and domain allowlists, and Tier 0 sandbox access. Grant permissions explicitly in **Capabilities**. Tool approval follows organization and tool policy; creating an agent does not grant blanket tool access.

## Agent review checklist

* Is the purpose narrow enough to test?
* Are the instructions clear about what the agent must and must not do?
* Are tools, knowledge sources, and domains limited to the use case?
* Has the agent been tested with representative and boundary-case inputs?
* Are the owner and approval path documented?


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