Solutions Engineers: Ready for MCP Demos?

Headless products are moving the SE's job from showing feature workflows on a screen to proving outcomes, or value engineering.

author
David Rhodes
published
2026-11-18
read
2 min

Last week’s HubSpot's AI product announcement about shifting to MCP, or "headless", represents an even bigger shift for SEs and product teams, because it also changes how demos happen.

On Nov. 13, HubSpot co-founder and CTO Dharmesh Shah posted that they are "leaning strongly into a headless version of HubSpot." While they already had a public MCP server; the shift was how much of the product they planned to expose. He said more of HubSpot's capabilities will be "consumable by AI agents and AI harnesses (like Claude and ChatGPT) via MCP." MCP, or “model context protocol” being the new standard for how AI assistants talk to other software platforms.

SaaS companies launching an MCP server is trending. In fact, Hubspot competitor Salesforce made their MCP announcement just over a year ago, and Nicolas Bustamante mused that "most SaaS should become an MCP server."

In a headless platform, the user doesn’t have to log on to your front end. They post questions to Claude or ChatGPT, which talk to your MCP for the answer…and that appears in the answer engine’s chat window.

For SEs, this trend creates two problems.

First, there’s the demo. SE’s control the screen today. They choose the dashboard, the workflow, and the order of the story, ideally based on customer discovery. With headless, the customer’s AI answer engine shows the result in its own front end. So…the SE no longer gets to plan what the buyer sees.

Second is measurement. Product analytics tools like Fullstory and PostHog track how users are navigating your app, the pages they view and the buttons they click. When an LLM talks to your product through MCP, that front end user journey is skipped entirely. Server logs still capture which tools were called; not the question the user asks Claude….or what the user does with the response.

My read: Headless products are moving the SE's job from showing feature workflows on a screen to proving outcomes, or value engineering. The most impressive demo moment becomes "ask ChatGPT this question and watch what data our MCP responds with." The SaaS product teams will excel by building MCPs an LLM can call easily, and they'll log finished workfows, not clicks.

Frequently asked questions

What does headless SaaS through MCP mean for customers?
In a headless SaaS platform, customers can ask questions through Claude or ChatGPT instead of logging into the SaaS product’s front end. The AI assistant communicates with the platform through Model Context Protocol (MCP), and the result appears in the AI assistant’s chat window.
What did HubSpot announce about headless access through MCP?
According to the post, HubSpot co-founder and CTO Dharmesh Shah said on November 13 that HubSpot was leaning strongly toward a headless version of HubSpot. HubSpot already had a public MCP server; the announced shift concerned exposing more HubSpot capabilities to AI agents and AI harnesses through MCP.
How does MCP change demos for solutions engineers?
In a traditional product demo, a solutions engineer chooses the dashboard, workflow, and presentation sequence. In a headless MCP interaction, the customer’s AI assistant presents the result in the AI assistant’s interface, reducing the solutions engineer’s control over what the buyer sees. The post argues that demos will shift toward proving outcomes through questions and the data returned through MCP.
What do product analytics miss when customers use a product through MCP?
The post explains that MCP interactions bypass the front-end navigation, page views, and button clicks tracked by tools such as Fullstory and PostHog. Server logs still capture tool calls, but those logs do not capture the customer’s question to Claude or what the customer does with the response.
What should SaaS product teams prioritize for headless MCP usage?
The post recommends building MCP capabilities that language models can call easily and logging completed workflows rather than clicks. The recommendation accompanies the author’s view that headless products shift solutions engineering toward outcome demonstration and value engineering.