# Create AI Agent DID

> ⚠️ **Before you begin...**
>
> Make sure you have configured the [MCP Client as per the setup instructions](https://docs.cheqd.io/product/getting-started/ai-agents/trust-registry/setup-mcp/..#getting-started).

After the MCP server is running and Claude Desktop (or any MCP Client) is connected, you can ask the agent to create a DID just by typing:

> Hello agent, can you create a DID on cheqd testnet?

{% hint style="success" %}
The agent may ask for human input to run the "create-did" MCP tool, please select "yes" or "allow".
{% endhint %}

<figure><img src="https://3569764573-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FPVAMvpKH7PYzvXA6u6Cn%2Fuploads%2FUkb4nQCmWproIR9Xf4Nk%2FScreenshot%202025-05-01%20at%2016.26.33.png?alt=media&#x26;token=359c01d6-3144-4e25-a519-2597a94a90bc" alt=""><figcaption></figcaption></figure>

## What is happening under the hood?

In the background, Claude is communicating with our MCP Server which relays a request to one of our SDKs, [Credo](https://docs.cheqd.io/product/sdk/credo). Using Credo, the AI Agent creates itself a DID that conforms to the [cheqd DID Method](https://docs.cheqd.io/product/architecture/adr-list/adr-001-cheqd-did-method).