HubSpot MCP: Connect Claude and ChatGPT to Your CRM

Most owners can use HubSpot's official connector inside Claude or ChatGPT without building a connection themselves. I would start with a question about the pipeline, then test one approved update before widening access. For the broader decisions behind using AI with your small business CRM, start with the job you want done.

What Is HubSpot MCP, and How Does It Differ from a Connector?

The remote MCP server is HubSpot's connection point for compatible AI tools. The official connectors put HubSpot access inside familiar chat applications. HubSpot lists connectors for Claude, ChatGPT, Gemini and Microsoft Copilot.

For most owners, the connector is the practical starting point. A custom client requires an MCP connector created in HubSpot, authentication using OAuth 2.1 with PKCE, and a connection to mcp.hubspot.com, according to HubSpot's developer documentation. HubSpot's separate local Developer MCP server is for building apps on its developer platform.

How Do You Connect HubSpot to Claude or ChatGPT?

I would make the first test small: read one known deal, check the answer against HubSpot, then approve one harmless update.

Connect HubSpot to Claude

1. Have a Super Admin approve the connector and select the permitted data access and eligible users.
2. In Claude, open Settings > Connectors, find HubSpot and connect.
3. Log in to HubSpot, choose the permissions to allow and confirm.
4. Return to chat and ask about a record you can check yourself.

These are HubSpot's documented setup steps. Claude requires an active HubSpot account, an active Claude account and a paid Anthropic subscription.

Connect HubSpot to ChatGPT

1. Check installation authority. Super Admins and users with App Marketplace Access can connect without prior approval; other users need Super Admin approval.
2. Follow HubSpot's ChatGPT connection guide to connect and authorise your account.
3. Configure the connector to always ask for approval before updates, as HubSpot recommends.
4. Test a CRM question. For deep research, click + > Deep research, then enable HubSpot under Apps.

Keep the initial permission set narrow enough that the person testing understands every proposed change.

What Can AI Read and Change in HubSpot?

As at 24 October 2026, HubSpot's availability announcement lists create and update access for CRM records including contacts, companies, deals and tickets, plus activities such as calls, meetings, notes, tasks and emails.

The Claude connector guide explicitly excludes deletes. Both connector guides cap bulk creates and updates at 10 records per operation. I would treat deletion as outside the connector workflow.

Revenue objects need care. HubSpot's announcement and developer documentation differ on read-only and beta objects, including quotes and line items. Check the current object table and your account's tools before designing around them. Do not assume invoices, quotes, orders or payments can be written.

The developer documentation also excludes joins, subqueries, association traversal and currency conversion. I would check the records behind a complex revenue answer before using the answer to make a decision.

What Does HubSpot MCP Cost, and Which Plans Do You Need?

The distinction is the connector charge versus the AI subscription. As at October 2026, this is what HubSpot publishes:

The connector guides say usage remains subject to HubSpot's API limits. Server availability does not establish unlimited usage. HubSpot's own AI features and credits are a separate purchasing question, covered in HubSpot AI agents and credits.

How Do Permissions, Sensitive Data and Australian Hosting Work?

Both connectors inherit HubSpot user permissions. That makes the user's existing access the starting point for review. Super Admin approval controls installation and access; it does not replace checking what each connected user can already see.

When Sensitive Data is enabled, the remote MCP documentation blocks activity and conversation data. The connector guides also block engagement data in that setting, and the ChatGPT guide excludes custom Sensitive Data Properties. A follow-up design that needs previous emails may therefore need a different approved route.

As at 24 October 2026, HubSpot lists an Australian data centre, but free-only accounts are assigned to the United States. Existing paid customers can migrate. Check your actual location rather than assuming an Australian business gets Australian hosting.

The connectors respect that location for data routed through HubSpot servers. Once data reaches Anthropic or OpenAI, those providers' terms apply. According to HubSpot's ChatGPT guide, connector data is excluded from training on Enterprise, Business and Education plans; consumer users can opt out in settings. Read the Claude guide's training terms and exceptions separately.

I would review permissions, write approval and data handling together. The guide to giving AI access to your inbox and CRM covers that decision.

When Is a Chat Connector Enough, and When Do You Need a Workflow?

A chat connector suits someone asking for a pipeline summary or requesting an update. The connector acts when someone asks. Repeatable follow-up needs a defined sequence, such as:

1. Trigger: an enquiry remains unanswered beyond the business's agreed interval.
2. Context: gather the contact, deal stage, owner and permitted emails or notes.
3. Draft: prepare a follow-up in the business's voice or a task for the owner.
4. Approval: a person checks the proposed message and record changes before action.
5. Record: log the outcome against the HubSpot record through a supported connection.

That is a workflow design, subject to the account's available access. The HubSpot lead follow-up guide develops the process.

For an Australian service business that wants repeatable HubSpot follow-up built and then run, with a person approving anything sensitive, Truespeak is built for exactly that. Before founding Truespeak, I spent more than 20 years in commercial roles, including e-commerce leadership at Optus, Samsung and Nikon.

Truespeak designs, builds and runs automation around the tools a business already uses. After launch, managed AI operations covers monitoring, fixing failures, reviewing exceptions and improving rules. Where a usable connection is unavailable, the step stays manual. I match the model to the job, so you do not need to choose Claude or ChatGPT first. The first step is a discovery call.