# AI Agents

> **Objective:** Create a GPU Pod on YottaLabs using RTX 5090, PyTorch, and JupyterLab through natural language with Cursor's MCP integration.

***

### Quick Start Guide

Follow these 5 steps to create and access your GPU Pod:

<details>

<summary>Step 1: Prerequisites Check</summary>

Before you begin, ensure you have the following:

**1.Node.js >= 18**

* **Check your version:** Run `node --version` in terminal
* **Not installed?** Download from [nodejs.org](https://nodejs.org/)

**2.Yotta API Key**

* Navigate to [Yotta Console](https://console.yottalabs.ai)
* Go to **Settings** → **Access Keys**
* Copy our apu keys securely

<figure><img src="/files/unTdbEdszt9A78a3pzZY" alt=""><figcaption></figcaption></figure>

**3.Cursor IDE**

* Download and install the latest version of [Cursor](https://cursor.sh/)

</details>

<details>

<summary>Step 2: Configure MCP Server</summary>

Configure the Yotta MCP server in Cursor to enable natural language GPU management.

**Locate Configuration File**

Find or create the configuration file based on your operating system:

| Operating System | Configuration File Path                    |
| ---------------- | ------------------------------------------ |
| **Windows**      | `C:\Users\<YourUsername>\.cursor\mcp.json` |
| **macOS**        | `~/.cursor/mcp.json`                       |
| **Linux**        | `~/.cursor/mcp.json`                       |

**Add Configuration**

Create or edit the `mcp.json` file with the following content:

```json
{
  "mcpServers": {
    "yotta": {
      "command": "npx",
      "args": ["-y", "@yottascale/agent-native-infra"],
      "env": {
        "YOTTA_API_KEY": "your-yotta-api-key-here"
      }
    }
  }
}
```

{% hint style="info" %}
**Important:** Replace `your-yotta-api-key-here` with your actual Yotta API Key from Step 1.
{% endhint %}

**Restart Cursor**

After saving the configuration:

1. **Completely close** Cursor (all windows)
2. **Restart** Cursor
3. The Yotta MCP server will now be available

</details>

<details>

<summary>Step 3: Create Your GPU Pod</summary>

Now you can create a GPU Pod using natural language through Cursor Composer.

**Method: Natural Language (Recommended)**

1. **Open Cursor Composer** by pressing:
   * Windows/Linux: `Ctrl+I`
   * macOS: `Cmd+I`
2. **Type your request** in natural language:

```
Create a Yotta Pod named pytorch-rtx5090-jupyter using RTX 5090 GPU, 
PyTorch image, enable JupyterLab, and set password to yotta2025
```

3. **Let AI handle it** - The AI will call the `pod_create` tool with appropriate parameters

**Key Parameters Reference ( just an example )**

<table><thead><tr><th width="155.24237060546875">Parameter</th><th width="112.6060791015625">Description</th><th>Example Value</th></tr></thead><tbody><tr><td><code>name</code></td><td>Pod name</td><td><code>pytorch-rtx5090-jupyter</code></td></tr><tr><td><code>image</code></td><td>Docker image</td><td><code>yottalabsai/pytorch:2.8.0-py3.11-cuda12.8.1-cudnn-devel-ubuntu22.04-2025081902</code></td></tr><tr><td><code>gpuType</code></td><td>GPU type</td><td><code>NVIDIA_RTX_5090_32G</code></td></tr><tr><td><code>gpuCount</code></td><td>Number of GPUs</td><td><code>1</code></td></tr><tr><td><code>environmentVars</code></td><td>Environment variables</td><td><code>JUPYTER_PASSWORD</code> (required)</td></tr><tr><td><code>expose</code></td><td>Ports to expose</td><td><code>22</code> (SSH), <code>8888</code> (JupyterLab)</td></tr><tr><td><code>regions</code></td><td>Preferred regions</td><td><code>["us-east-1", "us-east-2"]</code></td></tr></tbody></table>

</details>

<details>

<summary>Step 4: Wait for Pod Creation</summary>

After submitting your request, you'll receive a response confirming the pod creation.

**✅ Success Response Example**

```json
{
  "id": "420522713875330018",
  "name": "pytorch-rtx5090-jupyter",
  "gpuType": "NVIDIA_RTX_5090_32G",
  "gpuDisplayName": "RTX 5090",
  "gpuCount": 1,
  "singleCardVramInGb": 32,
  "singleCardPrice": "0.65",
  "status": "INITIALIZE",
  "environmentVars": [
    {"key": "JUPYTER_PASSWORD", "value": "yotta2025"}
  ],
  "expose": [
    {"port": 22, "protocol": "SSH"},
    {"port": 8888, "protocol": "HTTP"}
  ]
}
```

**Status Progression**

Your pod will go through these states:

1. **INITIALIZE** → Pod is being created
2. **RUNNING** → Pod is ready to use (usually takes 1-2 minutes)

</details>

<details>

<summary>Step 5: Access Your Pod</summary>

Once the pod status changes to **RUNNING**, you can access it.

**Via Yotta Console**

1. Open [Yotta Console](https://console.yottalabs.ai)
2. Navigate to **Compute** → **Pods**
3. Click on your pod: `pytorch-rtx5090-jupyter`
4. View the access URLs in the details panel

> Get the SSH address from the pod details in the console.

</details>

***

### Pod Management Commands

Once your pod is running, you can manage it using **natural language** commands in Cursor Composer:

#### List All Pods

```
List all my Yotta pods
```

#### Get Pod Details

```
Get details for pod pytorch-rtx5090-jupyter
```

#### Pause Pod (Stop Billing)

```
Pause pod pytorch-rtx5090-jupyter
```

{% hint style="info" %}
**Tip:** Pause pods when not in use to save costs!
{% endhint %}

#### Delete Pod

```
Delete pod pytorch-rtx5090-jupyter
```

{% hint style="info" %}
**Warning:** This action is irreversible!
{% endhint %}


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.yottalabs.ai/tutorials/ai-agents.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.