HubSpot Connector for ChatGPT: The Operator's Guide

The HubSpot connector for ChatGPT is HubSpot's official integration for OpenAI's ChatGPT, built on HubSpot's remote MCP server. It lets ChatGPT answer questions from your live CRM data, run deep multi-source research, and, since a February 2026 upgrade, create and update records with your approval. It works on every HubSpot plan at no extra cost; you need a paid OpenAI plan (Plus, Pro, Team, Enterprise, or Edu).

What is the HubSpot connector for ChatGPT?

It is HubSpot's official app inside ChatGPT, launched June 2025 as the first CRM deep research connector OpenAI shipped with any partner. Once connected, ChatGPT can answer everyday questions from your live portal - deal counts by stage, ticket volume by category - and switch into deep research mode for the kind of multi-variable analysis that used to need an analyst: cohort segmentation, churn pattern detection, territory planning.

It started as a read-only research tool. As of a Feb. 23, 2026 upgrade it also writes: creating and updating contacts, companies, deals, and tickets, logging calls, meetings, notes, tasks, and emails, straight from the ChatGPT chat window. That is a real capability jump, and it is also the single most confused fact about this connector right now, since plenty of marketing copy and marketplace listings written before that date still describe it as read-only. It is not, anymore.

Under the hood it runs on HubSpot's remote MCP server, the same plumbing HubSpot's connector for Claude is built on. The two are close cousins: same permission philosophy, same one-way-door admin quirks, different chat tool.

What can ChatGPT read and write in HubSpot?

Access is per object, and HubSpot's own permission table is the clearest source. Here it is, condensed:

Nothing can be deleted, on any object. Custom objects, unstructured data, and associated engagements are not yet available - HubSpot's docs say so directly, and it is the top complaint in OpenAI's own developer community from teams whose portals lean on custom objects.

How do you set up the HubSpot connector for ChatGPT?

Setup takes about five minutes, and the order matters:

1. Admin connects first. A Super Admin or a user with App Marketplace permissions opens ChatGPT, goes to Settings, then Apps, and connects HubSpot.
2. Choose the account and the permissions. Authenticate with HubSpot, pick the account to connect, and select which permissions the connector gets. Trimming permissions later means uninstalling and reinstalling, for the admin and every user, so decide the scope here.
3. Grant your team access. In HubSpot: Settings, then Integrations, then Connected Apps, then HubSpot connector for ChatGPT, then Give users access.
4. Each user connects their own ChatGPT account. They get only the permissions the admin selected, further limited by their own HubSpot user permissions.
5. Turn it on and set approvals. In ChatGPT, click the + icon, then More, then HubSpot (or just mention "HubSpot" in a prompt - it auto-enables). In the connector's tool settings, set Write tools to Needs Approval so nothing changes without a human confirming first.

If your team connected before Feb. 23, 2026, none of this replaces what you have; you upgrade instead. A ChatGPT workspace admin approves the updated app from a pop-up or the App Directory, then each user clicks Upgrade in the HubSpot app's settings and re-authenticates, making sure to select the new write and engagement permissions.

How do permissions and security work?

The core model: a Super Admin sets the outer boundary of what the connector can touch, and each user's own HubSpot permissions narrow it further. A rep sees in ChatGPT exactly the deals they can see in HubSpot, nothing more. Every create and update lands in the portal's Audit Log, attributed to both the user and the connector.

On data routing: requests go through HubSpot's data centers in your portal's region, so EU portals stay on EU routing. On training, OpenAI's policy actually splits by plan in a way worth knowing precisely - Enterprise, Business, and Education plans never use connector data for training. Free, Go, Pro, and Plus (the consumer tiers) can opt out in ChatGPT settings; if you do not, OpenAI may use the data to improve connector performance like tool use and retrieval accuracy, but says it will not use it for general model training. If your portal has Sensitive Data turned on, the connector gets no engagement data at all, by design.

Three one-way doors to know before you connect. Once a Super Admin turns on access for everyone, it cannot be turned off account-wide. Once users connect, admins cannot revoke the permissions they were given; narrowing scope means uninstall and reinstall for everyone. And uninstalling the connector does not disconnect your users; each has to uninstall individually. This is the identical shape of the same three doors on the Claude connector - HubSpot built both integrations on the same admin-control philosophy, and in both cases it is the thinnest part of an otherwise well-built permission model. Decide your scopes before the first connect, not after.

What are the limits to know before you connect?

The honest list, from HubSpot's own documentation and from what practitioners have actually hit:

• Custom validation rules are not applied. Pipeline stage validations and association label validations do not run on connector writes. ChatGPT can move a deal in a way the HubSpot UI would block. This is the single biggest thing to understand before granting write access.
• The bulk cap is a speed bump, not a guardrail. Bulk creates and updates are limited to 10 records per request. HubSpot's own best-practice advice, updating a single record before you go bulk, is worth following literally.
• No custom objects. Custom objects, unstructured data, and associated engagements are not yet available. A team on OpenAI's own developer forum hit this hard in March 2026: their custom objects were fully visible via HubSpot's direct API, just not through the connector.
• One HubSpot account at a time. A ChatGPT account connects to a single HubSpot account. Partners and agencies managing several portals disconnect and reconnect to switch.
• Published content only. For website pages, landing pages, and blog posts, only published data is available; drafts expose metadata only. To pull performance data for a specific page or post, you need to name the campaign and date range in your prompt.
• Web-only, for now. The connector runs in ChatGPT's web app. There is no desktop or mobile app support as of this writing, which is one real difference from the Claude connector's web, desktop, and mobile coverage.
• No deletes, anywhere. Honestly, a feature. ChatGPT cannot remove records, engagements, or property values on your behalf.
• User-reported: writes can fail silently in agentic flows. A HubSpot Community post from May 2026 describes ChatGPT's Agents/API path rejecting company property updates at the schema level, separate from the standard chat window HubSpot demos. Treat it as a reason to confirm a write actually landed in HubSpot before trusting an automated run, not as a documented HubSpot limitation.

How Hubjoy runs HubSpot through ChatGPT

We operate HubSpot portals through ChatGPT and Claude both, day to day: CRM operations, pipeline hygiene, reporting. This list comes from daily use, not a demo.

1. Scope the connection like a new hire. The admin sets the ceiling, the individual user's HubSpot access sets the floor. Connect with an account whose access matches what ChatGPT should see, not a Super Admin's view of everything.
2. Write tools on Needs Approval, permanently. Every create and update gets a human eye before it lands. Because validation rules do not run on connector writes, treat every write as if the UI guardrails were off. They are.
3. Reads before writes. Run it as a question-answering and reporting layer for a couple of weeks first. You learn which properties it reads reliably and where your data lies to it.
4. One record, then bulk. First writes go to a single low-stakes record. Widen to bulk only after you have confirmed changes land correctly, especially if anyone on your team is using ChatGPT's Agents or API surface instead of the plain chat window.
5. Check the Audit Log weekly. Connector writes are attributed to both the user and the connector, which makes a weekly filter cheap insurance against surprises.
6. Data hygiene still decides quality. ChatGPT answering from stale stages and duplicate contacts is confidently wrong. Clean the portal first; the connector amplifies whatever state your CRM is in.

The gotcha we plan around: same as the Claude connector, it is the one-way doors, not the write capability itself. Once you give your team access, you cannot take it back account-wide, and narrowing scope means everyone uninstalls and reinstalls. Five minutes of scope planning before the first connect saves a very awkward cleanup later. Our honest take, having run both: pick ChatGPT or Claude based on which chat tool your team already lives in day to day, not on which one is "more capable" - since the Feb 2026 upgrade, they do essentially the same CRM job.

Frequently asked questions

Need to connect a tool other than ChatGPT or Claude, like Cursor or a custom agent? Both packaged connectors run on HubSpot's own MCP server - see our guide to the HubSpot MCP server for the generic, developer-facing setup.

Connecting AI to your portal?

ChatGPT is only as good as the CRM it reads. We will tell you whether your portal is ready to be operated by AI, and what to fix first if it is not. Thirty minutes, no pitch.