On this page
What HubSpot supports natively
The answer to “is it read-only?” changed. HubSpot’s July 2025 launch of the ChatGPT deep research connector described “read-only access on core CRM objects” (contacts, companies, deals and tickets, with no engagements). The current HubSpot app goes much further, per HubSpot’s Knowledge Base article updated October 19, 2026:
| Access | Objects |
|---|---|
| Read, create and update | Contacts, leads, companies, deals, tickets, custom objects, line items, products, campaigns, landing pages, website pages, blog posts, marketing emails, segments (lists), pipelines, and calls, meetings, notes, tasks and emails |
| Read and create | Properties, marketing events, adding assets to campaigns |
| Read only | Users and teams, quotes, invoices, orders, carts, subscriptions, inbox and help desk conversations, campaign attribution, saved reports, AEO, buyer intent |
| Not supported | Deleting any object. Workflows and forms aren’t listed |
HubSpot’s article is inconsistent on custom objects: its table marks them readable and writable, while its FAQ says custom objects aren’t available yet. Test before you depend on them.
How to connect ChatGPT to HubSpot
- In ChatGPT, open Settings → Apps, select HubSpot and click Connect.
- Log in to HubSpot, choose the account, select the permissions to allow and click Connect App.
- If you aren’t a Super Admin or a user with App Marketplace access, a Super Admin approves the connector first and picks its optional data permissions.
- For a report, click +, choose Deep research, and turn HubSpot on under Apps (mentioning “HubSpot” in the prompt also turns it on). Citations link to the HubSpot records used.
Limits to plan around
- 10 records at a time for creates and updates.
- One HubSpot account per ChatGPT account.
- Set write tools to Needs Approval in the connector’s tool settings in ChatGPT. HubSpot warns that with Always allow, edits may occur without asking.
- Custom validation rules aren’t applied, including pipeline stage and association label validations, so a write can skip a rule your team relies on.
- Sensitive Data: if it’s on, ChatGPT can’t access engagement data, and custom Sensitive Data properties are never available.
- After a Super Admin changes permissions, each user reconnects.
Workflow drafts stay drafts: HubSpot says ChatGPT provides drafted workflow and email content for you to implement in HubSpot. If your question is setting up HubSpot rather than working its records, compare this with what Claude’s connector can do and the AI capability overview. To build properties, pipelines or workflows, we recommend Jetstack AI Projects over a chat connector; see how Projects compares with ChatGPT.
Where Jetstack AI fits
Jetstack AI’s MCP server connects to ChatGPT as a custom connector through developer mode, on plans that offer it: create a connector, enter https://mcp.jetstack.ai and choose OAuth. ChatGPT can then browse portals, import and deploy HubSpot assets with a reviewed mapping, and run audits within your Jetstack token’s scopes. Setup is in the MCP server docs; the developer overview covers the API and CLI.
That route copies setup that already exists. To build new setup, use Projects: describe what you need in a brief (for example, a deal pipeline with its required properties and a follow-up workflow), and Jetty, the Projects assistant, turns it into a plan you approve. It then generates the assets, deploys them only after a separate approval, and gives you a receipt checked by reading the assets back from HubSpot. Workflows are created switched off. Projects is currently in limited release; see what it can build.
Sources
- Set up and use the HubSpot connector for ChatGPT (HubSpot Knowledge Base)
- HubSpot connector for ChatGPT (HubSpot Developer Changelog, July 2025)
- Developer mode and MCP apps in ChatGPT (OpenAI Help Center)
- MCP server (Jetstack AI documentation)
Checked October 30, 2026. HubSpot changes its products often; tell us at team@jetstack.ai if something here is out of date.
AI, Claude & MCP
Do it from your own code or AI agent.
The Jetstack AI REST API, MCP server and CLI share API keys and scopes. Create and manage those keys in the app.
Read the developer docs