# MCP with ChatGPT

## How to Access a raia AI Agent via MCP in ChatGPT

This guide explains how to connect and use a **raia AI Agent** in ChatGPT using **Model Context Protocol (MCP)**.

You will learn how to:

* Enable Developer Mode
* Create a Custom MCP App
* Connect your raia MCP endpoint
* Publish it to your workspace
* Access and use the agent in ChatGPT

***

### Overview

ChatGPT supports external AI agents via **Model Context Protocol (MCP)**. By adding your raia MCP endpoint as a custom app, you can securely interact with your raia AI Agent directly inside ChatGPT.

⚠️ Note: Custom MCP servers are considered advanced configuration and may introduce risk if improperly configured. Only connect trusted endpoints.

***

## Step 1: Enable Developer Mode

1. Open **ChatGPT**
2. Go to **Settings**
3. Click **Apps**
4. Select **Advanced settings**
5. Toggle **Developer mode** ON

When enabled:

* You can add unverified MCP connectors
* Memory is disabled for that chat session
* You accept responsibility for the endpoint being added

<figure><img src="/files/GVGeKxDG28pUiASIPllx" alt=""><figcaption></figcaption></figure>

<figure><img src="/files/kO1nJ9r8J29yVSZKNNHR" alt=""><figcaption></figcaption></figure>

***

## Step 2: Create a New MCP App

1. In **Settings → Apps**
2. Click **Create app**

You will see the **New App (Beta)** screen.

Fill in the following:

#### App Configuration

| Field              | What to Enter                                                        |
| ------------------ | -------------------------------------------------------------------- |
| **Name**           | `raia DIS` (or your preferred agent name)                            |
| **Description**    | Optional                                                             |
| **MCP Server URL** | Your raia MCP endpoint (e.g. `https://api.raia2.com/mcp/{agent-id}`) |
| **Authentication** | Select **No Auth** (if your endpoint does not require OAuth)         |

Example MCP URL format:

```
https://api.raia2.com/mcp/{agent-uuid}
```

<figure><img src="/files/Pbo5ZHELr40LaeNiytVS" alt=""><figcaption></figcaption></figure>

<figure><img src="/files/fSajVSWywAPgVt3qYwlC" alt=""><figcaption></figcaption></figure>

***

### Step 3: Accept Risk Acknowledgment

Before creating the app, you must confirm:

> "I understand and want to continue"

This acknowledges:

* OpenAI has not reviewed the MCP server
* The endpoint could access or modify data
* You trust the MCP server being connected

Check the confirmation box and click **Create**.

***

## Step 4: Verify App Connection

After creation, you will see the app listed under:

**Settings → Apps → Enabled apps**

You should see:

* Your app name (e.g. `raia DIS`)
* A `DEV` badge (indicating developer mode)
* Connection status
* Authorization used (None if No Auth)

You can:

* Disconnect
* Delete
* Publish to workspace

***

## Step 5 (Optional): Publish to Your Workspace

If you want others in your workspace to use the raia agent:

1. Click the app
2. Click the **••• menu**
3. Select **Publish**

You will be prompted to:

* Review potential risk: Unauthorized data access
* Confirm understanding
* Optionally configure action confirmations

Click **Publish** to make it available to workspace members.

After publishing, it will appear in the **Apps Directory** for your workspace.

<figure><img src="/files/70NvtQ1bRLMlYd3bn5dS" alt=""><figcaption></figcaption></figure>

<figure><img src="/files/tnwLOWULK5Pokm4ICcKT" alt=""><figcaption></figcaption></figure>

***

## Step 6: Use the raia Agent in Chat

Once enabled:

1. Open a new ChatGPT conversation
2. Type `@`
3. Select your raia agent (e.g. `@raia DIS`)

You can now interact with your raia AI Agent directly within ChatGPT.

Example:

```
@raia DIS Get account summary for customer 12345
```

The MCP connection allows ChatGPT to:

* Send structured requests to your raia agent
* Execute defined MCP actions
* Return responses in-chat

***

## How It Works (Technical Summary)

* ChatGPT connects to your MCP endpoint via Server-Sent Events (SSE)
* The raia MCP server exposes defined actions
* ChatGPT calls those actions when prompted
* Responses are streamed back into the conversation

If Developer Mode is enabled:

* Memory is disabled
* The session does not retain long-term context

***

## Security Considerations

When connecting MCP endpoints:

✔ Only use trusted HTTPS endpoints\
✔ Restrict access where possible\
✔ Audit usage periodically\
✔ Avoid exposing sensitive actions without confirmation

When publishing to workspace:

* Limit who can access the app
* Configure action confirmations for consequential operations

***

## Troubleshooting

#### App does not appear in chat

* Confirm Developer Mode is enabled
* Verify the app is listed under Enabled Apps
* Refresh ChatGPT

#### MCP connection fails

* Verify the endpoint URL is correct
* Confirm the MCP server is reachable
* Check SSL configuration

#### Actions not triggering

* Confirm the MCP server exposes properly defined actions
* Ensure the prompt clearly references the agent using `@agent-name`

***

## Summary

By connecting your raia AI Agent via MCP in ChatGPT, you enable:

* Direct AI-agent interaction inside ChatGPT
* Action execution against raia systems
* Workspace-wide deployment (optional)
* Secure, extensible agent integration

This allows raia agents to operate seamlessly within ChatGPT’s conversational interface while maintaining control through MCP.


---

# 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.raiaai.com/integrations/tools-integration/mcp-with-chatgpt.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.
