HubSpot yesterday published a strategic vision from Duncan Lennox, its Chief Product and Technology Officer, describing the company's plan to turn its customer platform into open infrastructure that any AI agent - not just HubSpot's own - can read, write, and act upon. The post, titled "Our Vision for Building an Open Ecosystem for the Agent Era" and updated May 4, 2026, lays out an architectural direction that separates HubSpot from CRM and marketing platforms taking the opposite approach.

The core argument is deceptively simple. According to Lennox, agents do not click through dashboards or navigate interfaces. They call APIs, read structured outputs, and take action. That distinction has direct consequences for any software vendor that has historically locked its capabilities behind a graphical interface: if agents cannot reach those capabilities programmatically, those capabilities become inaccessible to the growing layer of automation that is taking over go-to-market workflows.

What HubSpot is opening

According to the blog post, HubSpot is opening two distinct layers to external agents, partners, and developers.

The first is what Lennox calls the data layer: contacts, companies, deals, conversations, tickets, and activity records. This is not new territory. Thousands of integrations already pull data from HubSpot's CRM through its existing APIs. What the post signals is a commitment to keeping that layer open and extending it, rather than restricting access as AI introduces new forms of automated consumption. According to Lennox, bringing data into HubSpot remains free, and a customer's data remains theirs. If they choose to leave, it goes with them.

The second is the intelligence layer, which is the more significant of the two. This covers outputs that HubSpot generates from its data - scores, assessments, and benchmarks that have historically been available only inside HubSpot's own interface. According to the post, that will change. Agents outside HubSpot will be able to call those intelligence outputs directly through an API. The specific examples Lennox gives include a deal risk score, a lead qualification signal, and a ticket resolution recommendation. These are not raw data points - they are derived signals built on patterns observed across HubSpot's network of more than 280,000 customers.

The deal intelligence example is the most technically specific in the post. Lennox describes a scenario where a sales manager pulls open pipeline data - amount, stage, close date, last activity - into a large language model and asks what is at risk. The model can calculate averages from the numbers in front of it, but it does not know whether 30 days in-stage is fast or slow for a given industry. It does not know that the champion on one deal went quiet after a reorganization. It does not know that a comparable company stalled on the same objection last quarter.

According to Lennox, a single API call to HubSpot's intelligence layer will return a pre-computed risk score built on patterns across hundreds of thousands of customers. It will surface the industry benchmark, flag the champion's silence, and identify the pattern match to prior stalled deals. It will also suggest a next step - flagging the deal for review or triggering a follow-up - rather than leaving the reasoning to the agent or the user. This is what Lennox means when he describes the intelligence layer as giving an agent a "head start" that no standalone model can replicate.

The MCP server and what is live today

According to the post, HubSpot's APIs and MCP server are already live. Connectors for Claude, ChatGPT, Gemini, and Copilot are active and delivering value to customers. More than 2,000 apps run across the HubSpot ecosystem, and new agents are being built on top of the platform every week, according to Lennox.

HubSpot was among the early CRM vendors to ship an MCP server. The company launched its first CRM connector for Anthropic's Claude on July 29, 2025, enabling customers to access their relationship management data directly within Claude through natural language queries. That followed the ChatGPT deep research connector announced on June 4, 2025, which made HubSpot the first CRM platform to integrate with OpenAI's platform for advanced data analysis.

The Model Context Protocol has become the dominant technical standard for connecting AI agents to external platforms across the marketing technology industry. As PPC Land has documented, AdRoll and PubMatic demonstrated MCP-powered agent-to-agent diagnostics on April 23, 2026, allowing a demand-side agent to query supply-side deal diagnostics in real time without switching platforms. Amazon opened its advertising APIs to AI agents through MCP in open beta in February 2026. Meta followed with open beta AI connectors for Claude and ChatGPT on April 29, 2026. HubSpot's announcement today fits directly within this pattern: the major platforms are systematically dismantling the access barriers that previously required custom API development and ongoing engineering maintenance.

Full API parity as the stated target

The most consequential commitment in the post is what Lennox calls full API parity. According to the blog, HubSpot is continuing to expand its public API surface so that every capability of the platform - every workflow, every action, every piece of context - is accessible to apps and agents. The stated principle is explicit: no capability should live only behind a UI.

That is a significant operational target. Many enterprise software platforms have grown over years through acquisition and product expansion, leaving large portions of their functionality accessible only through interfaces that were designed for human navigation. Reaching API parity across a platform of HubSpot's breadth - covering marketing, sales, and service workflows, plus the intelligence layer being built now - represents sustained engineering investment, not a single release.

The post does not provide a timeline for when full API parity will be achieved. It positions it as a direction the company is committed to, with the current state described as "coming next" rather than complete. What is complete today, according to Lennox, is the foundational layer: the MCP server, live connectors to the four major AI platforms, and the existing API surface that already powers thousands of third-party integrations.

How HubSpot frames the competitive logic

Lennox addresses the strategic choice directly in the post. According to him, some platforms will respond to the shift toward agents by closing down, constructing walled gardens, restricting access, and making it harder for customers to benefit from AI. HubSpot is making the opposite bet.

The framing rests on three stated principles. Customer value comes first: HubSpot says it will always invest in its own first-party agents through its Breeze toolset, but acknowledges that the best agent for a specialized industry or workflow will often come from the ecosystem. Open by design means anything that can be done inside HubSpot should be doable through an API - and HubSpot's intelligence should be accessible wherever people work, not only inside HubSpot's own interface. Trusted by default means governance is treated as core infrastructure: when a customer connects a partner tool, spins up an agent, or builds something custom, they should know exactly what it can access and what it is doing.

The governance point is worth noting separately. As PPC Land has covered, the IAB Tech Lab Agent Registry reached 10 active entries in March 2026, all operating under MCP, precisely because trust and auditability have become genuine infrastructure requirements as agents begin acting on behalf of businesses. HubSpot's framing of trust as a design principle rather than an afterthought aligns with the direction the broader industry governance layer is moving.

The 280,000-customer network as a structural advantage

The intelligence layer argument depends entirely on scale. A risk score built on patterns from one company's pipeline is a calculation. A risk score built on patterns from 280,000+ companies, segmented by industry, deal size, sales cycle length, and objection type, is a benchmark. The difference is not marginal.

According to Lennox, this network scale - built across two decades of operating a CRM platform - is what separates HubSpot's intelligence outputs from what a standalone model can produce. A large language model reasoning over raw CRM records has no way to know what is normal for a specific business, or what has worked for hundreds of thousands of companies like it. The intelligence layer is HubSpot's answer to that gap: pre-computed context that agents can call rather than derive.

The same logic applies to the Breeze agents that already operate inside HubSpot. According to the post, the work those agents do - qualifying leads, resolving tickets, saving deals - will soon be available to external teams and agents wherever they operate, not only within the HubSpot interface.

Why this matters for the marketing community

The shift Lennox is describing has direct implications for how marketing and sales teams build their technology stacks. The traditional CRM evaluation was largely about which platform had the best interface, the best reporting, and the most integrations. The emerging evaluation is different: which platform exposes its intelligence to the agents that are increasingly doing the work?

HubSpot's AEO tool, launched on April 14, 2026 as part of the company's Spring 2026 Spotlight, showed how the company is responding to the structural decline in organic search traffic - down 27% year-over-year for HubSpot customers according to the company's own data. The open ecosystem announcement is a complementary move: if agents are increasingly where go-to-market work happens, then the platform that powers that work needs to be where agents can reach it.

The broader pattern across the marketing technology industry, as PPC Land has tracked through 2025 and into 2026, is consistent with what Lennox is describing. Platforms that built their value on proprietary interfaces are under pressure. The ones building toward open API access and MCP compatibility are positioning themselves as infrastructure rather than destinations. Whether HubSpot can deliver full API parity across its entire product surface remains to be seen - but the architectural direction it has stated today puts it clearly in the open camp at a moment when that choice has real competitive weight.

Timeline

Summary

Who: Duncan Lennox, Chief Product and Technology Officer at HubSpot, published the announcement. HubSpot, Inc. (NYSE: HUBS) is the company behind the platform, which serves more than 280,000 customers globally across marketing, sales, and service functions.

What: HubSpot today laid out a vision for making its entire platform accessible to external AI agents through two open layers - a data layer covering contacts, companies, deals, conversations, and tickets, and an intelligence layer covering pre-computed scores, assessments, benchmarks, and action recommendations built on patterns from its customer network. The company confirmed that its MCP server and APIs are already live, with connectors active for Claude, ChatGPT, Gemini, and Copilot. It committed to full API parity as the next milestone, meaning every platform capability will eventually be accessible programmatically without a graphical interface.

When: The blog post authored by Duncan Lennox was published and updated on May 4, 2026.

Where: The announcement was published on HubSpot's blog at hubspot.com. The platform infrastructure it describes - the MCP server, APIs, and connectors - operates globally, accessible from any environment where customers, partners, and developers build agents and integrations.

Why: According to Lennox, agents do not navigate interfaces - they call APIs. Software built for human interaction must evolve to be genuinely accessible to agents. HubSpot's strategic argument is that openness, not restriction, is the right response to a moment when agents can access data, act on behalf of customers, and run business processes. The intelligence layer, built on two decades of CRM data across 280,000+ businesses, is positioned as the specific asset that gives HubSpot's open platform a structural advantage that standalone models and closed platforms cannot replicate.

Share this article
The link has been copied!