> ## Documentation Index
> Fetch the complete documentation index at: https://docs.growthxai.com/llms.txt
> Use this file to discover all available pages before exploring further.
# GrowthxAI Outreach Plugin: Connect ChatGPT to Your Workspace
> Triage replies, build campaigns and check sender health by chatting with ChatGPT.
The **GrowthxAI Outreach** plugin lets ChatGPT work in your outreach workspace. It has the same tools as the [Claude connector](/integrations-api/mcp-connect-claude), so you can ask ChatGPT things like:
* *"Any pending replies?"*
* *"Build a 4-touch sequence for fintech CFOs and dry-run it on the Acme list."*
* *"Why isn't anything sending?"*
* *"Give me a client report for last month."*
ChatGPT works with **exactly your role and client scope**: it can't see or do anything you couldn't do in the app.
## How to add the plugin
Adding a plugin by its URL needs **Developer mode**, which is available on paid ChatGPT plans. On a Business or Enterprise workspace, an admin may have to allow it first. If you can't find the setting, ask your workspace admin.
1. In ChatGPT, open **Settings → Security and login** and turn on **Developer mode**.
2. Go to [chatgpt.com/plugins](https://chatgpt.com/plugins) and click the **+** button.
3. Name it **GrowthxAI Outreach**, and under **Connection** paste this URL:
```text theme={null}
https://ktwqkvjuzsunssudqnrt.supabase.co/functions/v1/outreach-mcp/mcp-chatgpt
```
4. Choose **OAuth** as the authentication, then create the plugin.
5. Sign in with your GrowthxAI account and approve access.
Start a **new chat**, add **GrowthxAI Outreach** from the tools menu, and ask *"What's in my outreach workspace?"*
Use the URL ending in `/mcp-chatgpt` for ChatGPT. The Claude URL, which ends in `/mcp`, won't sign in from ChatGPT.
If you belong to several workspaces, ChatGPT asks which one to use.
## What you can do
It's the same list as in Claude:
| Area | Examples |
| - | - |
| **Inbox triage** | *"Any pending replies?"* ChatGPT reads every conversation waiting on you, drafts replies, and shows them in one table for you to accept, edit or skip. |
| **Campaigns** | *"Build a sequence for …"* It checks sender capacity, writes the steps, validates them, estimates the timing, and enrolls leads once you approve. |
| **Leads** | Clean up and import a list, tag leads, set stages, add suppressions |
| **Sender health** | *"How are my senders doing?"* Health, capacity and the reasons behind any problems |
| **Diagnosis** | *"Why isn't sequence X sending?"* Plain-language causes, such as a daily cap, working hours, warm-up or a disconnected sender |
| **Reports** | Overview, funnel, reply intent, sequence, sender, client, cost and deliverability: the same numbers as the [Reports page](/reports/reports-overview) |
| **Safe editing** | Save a draft, see who a publish would affect, publish, move pinned leads forward, rebalance a pool |
| **Recovery** | *"Which leads failed, and why?"* then retry, skip or exit them |
| **AI lines** | Write per-lead opening lines for a list, then show them to you for approval |
The full list of tools is in the [functions reference](/integrations-api/mcp-functions-reference).
## How ChatGPT differs from Claude
* **No ready-made prompts.** Claude Desktop has slash-command prompts such as *Morning inbox triage*. ChatGPT doesn't show them, so just ask in plain language. The plugin carries the same instructions either way.
* **Importing a list.** ChatGPT can read rows you paste or attach in the chat and send them in batches. It can't pass an attached file straight to the plugin, so for a large list, upload the CSV in the app's [importer](/crm-lead-management/importing-leads) and ask ChatGPT to check the result.
* **Extra approval prompts.** ChatGPT may show its own approval prompt before a tool changes something. That is separate from GrowthxAI's own confirmation, described next.
## ChatGPT always asks before acting
Actions that send, enroll, suppress, import, activate, restore or export are **confirmation-gated**:
1. ChatGPT first shows you a summary of exactly what will happen, for example *"Send 6 replies from 3 senders"*.
2. Nothing happens until you say **yes**.
3. Your approval is valid for 10 minutes and only for that exact action. If anything changes, ChatGPT has to ask again.
If a prospect writes again after ChatGPT drafted a reply, the send is skipped so you don't answer an old message.
## What ChatGPT can't do
* Raise daily limits, change working hours, or connect and disconnect senders
* Manage members or billing
* Send LinkedIn messages or invitations outside a sequence or an existing conversation
* Get around blacklists, caps or reply-stop
* **Send an AI-written line nobody approved.** It can write them and show them to you; approving is yours.
## Usage limits
Per person, per hour: 600 reads, 120 changes, 60 bulk actions and 20 confirmed actions. Per day: 300 sends and 500 AI drafts. The limits are counted per person, so they're shared between Claude and ChatGPT. If ChatGPT says it's being rate-limited, wait a little and try again.
## Q: I can't sign in, or the plugin stopped working
* Check the URL ends in `/mcp-chatgpt`.
* Open **GrowthxAI Outreach** at [chatgpt.com/plugins](https://chatgpt.com/plugins) and reconnect it. If that doesn't help, remove the plugin and add it again with the URL above.
* Start a new chat afterwards, and add the plugin from the tools menu again.
Any drafts that weren't sent are still in your chat, so nothing is lost.
## Related articles
* [GrowthxAI Outreach Connector: Connect Claude to Your Workspace](/integrations-api/mcp-connect-claude)
* [GrowthxAI Outreach MCP: Complete Functions Reference](/integrations-api/mcp-functions-reference)
* [The Public API](/integrations-api/public-api): for unattended automations, where a plugin needs a person
* [Inbox Overview](/inbox-and-manual-messaging/inbox-overview)
* [Roles and Permissions](/getting-started/platform-overview/roles-and-permissions)