---
title: "ChatGPT and OpenAI agents + OmniChannels.ai: send and track messages"
description: "Give OpenAI agents built with the Responses API or Agents SDK our MCP server so they can send WhatsApp, SMS and RCS messages and read delivery reports."
canonical: "https://omnichannels.ai/integrations/chatgpt/"
language: "en"
updated: "2026-10-10"
publisher: "BroadNet Technologies"
brand: "OmniChannels.ai"
---

# ChatGPT and OpenAI agents + OmniChannels.ai: send and track messages

Canonical: https://omnichannels.ai/integrations/chatgpt/
Language: en
Publisher: OmniChannels.ai by BroadNet Technologies

**In short:** OpenAI agents connect to OmniChannels.ai through our remote MCP server at https://mcp.broadnet.me/mcp. The Responses API takes it as a remote MCP tool with your token in an Authorization header, and the Agents SDK connects over Streamable HTTP. The agent can then send WhatsApp, SMS and RCS messages, use approved templates and read delivery reports. There is no ChatGPT app to install.

Connect via MCP · Connect via API & webhooks

## What you can do

### Customer updates from your own agent

An agent you build on the Responses API sends order or appointment updates when your app asks it to.

> Try asking: "Send a WhatsApp message to 971500000000 saying their order has shipped and will arrive tomorrow."

### Choose and fill approved templates

The agent lists your approved WhatsApp templates and fills the variables of the one that fits.

> Try asking: "List my approved Arabic WhatsApp templates and send order_update to 966500000000 with order number INV-400."

### Delivery questions for your team

A read-only agent with only the delivery report tool answers questions about failed or pending messages.

> Try asking: "How many messages in yesterday's statements campaign failed, and what were the most common reasons?"

### Draft first, send after review

Require approval for send tools so a person checks the copy and recipients before anything goes out.

> Try asking: "Draft an SMS reminder in English and Arabic for tomorrow's appointments and show it to me before sending."

### Combine with your business data

When the agent also has a tool or MCP server for your store or CRM, it can fetch the recipients there and send through OmniChannels.ai.

> Try asking: "Send a WhatsApp order update to customers whose orders shipped today."

## How it connects

OmniChannels.ai runs a remote MCP server over Streamable HTTP. Every request carries your API token in an Authorization: Bearer header, and each tool acts with the permissions, sender IDs and balance of that token, exactly as your own code would.

In the OpenAI Responses API you add an MCP tool with server_url set to our endpoint and the Authorization header in headers, and you can require approval before tool calls. In the Agents SDK, MCPServerStreamableHttp takes the same URL and headers. ChatGPT itself can add remote MCP servers as connectors in developer mode on plans that include it, but its connector settings may not accept a fixed bearer header; check OpenAI's current documentation.

- Endpoint: https://mcp.broadnet.me/mcp (Streamable HTTP)
- Authentication: Authorization: Bearer <token>; use a temporary key while testing
- Responses API: remote MCP tool with server_url and headers
- Agents SDK: MCPServerStreamableHttp with url and headers
- Without MCP, an agent can call our REST API using the OpenAPI spec

## Set it up

1. **Request API access:** Request test access and receive an API token. Test sends go only to your own verified numbers.
2. **Get an MCP key:** In the admin portal open Integrations, then MCP Integration, and copy a temporary key. (Guide: https://omnichannels.ai/developers/mcp/)
3. **Attach the server to your agent:** Add an MCP tool with server_url https://mcp.broadnet.me/mcp and an Authorization header, or create MCPServerStreamableHttp in the Agents SDK, reading the token from an environment variable.
4. **Limit tools and require approval:** Allow only the tools the agent needs and require approval for every tool that sends, since each send is a real, billed message.
5. **Test with your own number:** Ask the agent to send a test message to your verified number, then check the delivery report. (Guide: https://omnichannels.ai/developers/webhooks/)

## Example workflow

OpenAI agent (Built with the Responses API or Agents SDK) → OmniChannels.ai MCP server (Send tool runs with your token after approval) → Customer's phone (WhatsApp, SMS or RCS message delivered) → Delivery report (The agent reads the status back with a read tool)

## Questions

### Is there an official OmniChannels.ai app in ChatGPT?

No. OpenAI agents connect through OpenAI's standard support for remote MCP servers, using our public MCP endpoint and your API token. There is nothing to install from an app directory.

### Can I add the server to ChatGPT directly?

ChatGPT can add remote MCP servers as connectors in developer mode on plans that include it. Our server expects a fixed Authorization: Bearer header, which ChatGPT's connector settings may not support, so the Responses API and Agents SDK are the reliable routes. Check OpenAI's current documentation for your plan.

### Can the agent send without approval?

That depends on how you configure it. The Responses API lets you require approval for MCP tool calls; keep it on for every send tool and allow only the tools the agent needs.

### Does OpenAI get access to my customer data?

The model sees what the tools return, such as templates and delivery reports, plus whatever your application puts in the conversation. The token limits the agent to your account's permissions.

## Developer guides

- https://omnichannels.ai/developers/mcp/
- https://omnichannels.ai/developers/openapi/
- https://omnichannels.ai/developers/quickstart/
- https://omnichannels.ai/developers/authentication/

## Vendor documentation

- [Remote MCP — OpenAI platform documentation](https://platform.openai.com/docs/guides/tools-remote-mcp)
- [Model Context Protocol — OpenAI Agents SDK](https://openai.github.io/openai-agents-python/mcp/)
- [MCP and connectors — OpenAI platform documentation](https://platform.openai.com/docs/mcp)

## Related integrations

- https://omnichannels.ai/integrations/claude/
- https://omnichannels.ai/integrations/google-gemini/
- https://omnichannels.ai/integrations/langchain/
- https://omnichannels.ai/integrations/n8n/

ChatGPT and OpenAI Agents is a trademark of its owner. OmniChannels.ai is not affiliated with or endorsed by ChatGPT and OpenAI Agents; this page describes a connection built with ChatGPT and OpenAI Agents's own MCP, HTTP or webhook features.

Updated: 2026-10-10