HubSpot published a pricing post on November 17 that sets out how it will charge for headless access, the arrangement in which AI tools such as Claude and ChatGPT work inside a CRM without a person opening the CRM screen. The company splits the answer into three tiers: seat-included connector use, API calls from agents a customer brings, and governed agent work that draws down HubSpot Credits.

What HubSpot says it will charge for

The post, written by Duncan Lennox and updated on November 18, opens with the first tier: every HubSpot seat comes with enough headless access to complete everyday work from Claude or ChatGPT at no additional cost. HubSpot lists connectors for Claude, ChatGPT, Gemini and Microsoft Copilot, plus its MCP server and Agent CLI, and says a sales rep can summarize open deals from them, a customer success manager can prepare for a renewal call, and a marketer can launch a campaign without opening another tab.

The second tier covers agents a customer has built or bought. HubSpot says they connect through the same APIs that thousands of integrations already use, that the calls count toward the API capacity already included in the plan and that they do not consume credits. For teams that exceed that capacity, HubSpot says additional capacity can be added at any time.

Advertisement

Simplified Management — Advertisement

The third tier is where credits apply. HubSpot says it is building a governed environment for agents, with purpose-built agentic APIs that carry deeper business context. Customers get control over what each agent can access and do, a record of every action and visibility into which agent did what. Work that agents do at scale in that environment, such as enriching records and personalizing outreach, “is paid for with credits,” in the post’s words. HubSpot also says power users who lean on AI tools more than most can use credits to expand the capacity included with their seat, without buying an add-on or an extra seat type.

The post frames the choice against the rest of the market. HubSpot writes that many platforms restrict AI access to their most expensive tier, sell it as an add-on or charge for the same work twice, and it does not name them. Its own position, in its words: “your data belongs to you.”

Where this sits next to Salesforce’s headless push

HubSpot’s post lands three months after Salesforce expanded its own headless program. In an August 19 announcement, Salesforce introduced a Headless 360 MCP Server that lets agents running in Agentforce, Claude, ChatGPT, Cursor and other AI platforms discover and invoke Salesforce capabilities, and a Data 360 MCP Server that exposes nearly 200 Data 360 APIs. It also made the Slackbot MCP Client generally available, connecting Slackbot to Salesforce MCP servers and more than 20 partner applications, including Atlassian, Box, Canva, Docusign, Notion and Zoom.

Salesforce’s announcement describes what agents can reach and how identity, permissions and governance carry over. It does not address what agent access costs, and it carries a general note that pricing and packaging are subject to change. The two documents answer different questions, which is why a buyer cannot compare them line by line: one is an architecture announcement, the other a price structure.

Zoom’s Chief Ecosystem Officer, Brendan Ittelson, said in the Salesforce release: “Zoom’s open platform strategy is built on meeting customers where they work.” The sentence describes the same shift HubSpot is pricing, with the work happening in the AI tool the seller already has open and the CRM acting as the system underneath.

Two earlier pieces on this site give context. We wrote about how the press read HubSpot’s move to headless CRM and about KeyBanc’s finding on software firms that charge separately for AI. Both stories were about where the money goes once AI work moves into the CRM.

Newsletter

Get the week's best tech coverage.

Free. Read by thousands of HR, tech, and business leaders.

What it means for the sales leader

In our reading, the useful part of the post is the boundary it draws between reading a CRM and acting inside it at volume. Reading and light updates from a rep’s own Claude or ChatGPT session fall inside the seat. Agents that enrich records or personalize outreach across a whole book of accounts fall into the credit meter. A forecast for next year’s CRM spend therefore depends on how many of your agents belong in the second group.

Three checks belong in the next renewal conversation with any CRM vendor:

  • Which agent actions are covered by the seat, which count against API capacity and which consume credits, written into the order form rather than a blog post.
  • Whether API capacity includes calls from third-party agents, and what happens at the limit.
  • Whether the audit trail HubSpot describes, a record of every action and which agent did what, is available for agents that connect through plain APIs or only inside the governed environment.

The practical gap between the tiers is volume and oversight. A rep asking Claude to summarize ten open deals is one kind of use. A prospecting agent that personalizes outreach across a large contact list is another, and HubSpot’s post places that second kind in the governed environment. Finance teams that budget CRM by seat will find that the seat count no longer predicts the whole bill.

Sales operations teams can also start measuring now. Log which workflows your reps already run from Claude or ChatGPT, and which agents you plan to run unattended. The first list tells you what the seat should cover. The second tells you where credit consumption will come from, and gives you a number to take to the vendor before the contract is signed.

Source: HubSpot, “HubSpot is headless. Here’s how we’ll price it”