For years, HubSpot invested in making our platform the most productive position for advertising, gross sales, and repair groups to do their paintings. With AI, we’ve been development it to do the paintings for them – thru brokers that qualify leads, get to the bottom of tickets, save offers, and power results around the trade. That’s why we name HubSpot an agentic buyer platform.
However brokers don’t click on thru dashboards or navigate interfaces; they name APIs, learn structured outputs, and take motion. Tool constructed for people has to adapt to be in truth available to brokers, too.
Get right of entry to on my own isn’t sufficient, even though. Brokers additionally want substance. An agent reasoning over uncooked information has no method to know what’s standard for a particular trade, or what’s labored for masses of 1000’s of businesses find it irresistible. As we not too long ago wrote, the actual AI race isn’t about fashions or knowledge; it’s about context.
That conviction shapes the entirety we construct. It’s why we have been a number of the first to send an MCP server, and why we’ve stored increasing what brokers can learn, write, and act on since. That was once handiest the start.
The imaginative and prescient we’re running towards is larger: Brokers can run on HubSpot. And brokers can run HubSpot.
Operating on HubSpot manner any agent – ours or someone else’s – can plug into HubSpot’s knowledge, context, and features as a development block. Operating HubSpot manner brokers can perform the platform end-to-end thru our APIs, MCP server, CLI, and no matter get admission to strategies come subsequent.
Contents
What we’re opening up
For brokers to run on HubSpot and to run HubSpot, they want what we name expansion context. That’s the particular, dynamic figuring out AI must ship actual effects around the go-to-market, making an allowance for the entirety about an organization’s trade, groups, processes, and consumers – and strengthened by way of patterns throughout HubSpot’s community of 280,000+ consumers.
It’s derived from two issues – knowledge and intelligence – and we’re opening each to our ecosystem of shoppers, companions, and builders.
The knowledge layer is the root: contacts, firms, offers, conversations, tickets, process – open and available, powering 1000’s of integrations nowadays. As all the time, bringing knowledge into HubSpot is loose. A buyer’s knowledge is theirs. In the event that they ever make a choice to depart, it is going with them.
The intelligence layer is what we’re development now. It covers each insights that tell selections (rankings, tests, and benchmarks that may be known as immediately) and movements that power results (qualifying leads, resolving tickets, saving offers). That is the paintings our Breeze brokers already do inside of HubSpot, and it’ll quickly be to be had anywhere groups and brokers perform.
As an example, take deal intelligence. A gross sales supervisor pulls their crew’s open pipeline into an LLM – quantity, level, shut date, ultimate process – and asks what is in peril. The type can calculate averages from the knowledge in entrance of it, but it surely does not know whether or not 30 days in-stage is speedy or gradual for this business. It does not know the champion on any such offers went quiet after a reorg. It does not know a identical deal at a related corporate stalled on precisely this objection ultimate quarter.
With the intelligence layer, a unmarried API name will go back a pre-computed chance rating constructed on patterns throughout HubSpot’s masses of 1000’s of shoppers. It’s going to know that this business’s gross sales cycle runs 90 days, no longer 30. It’s going to know the champion went quiet after a reorg. It’s going to know that different offers like this one stalled at the similar objection prior to. And it’ll be capable of act on that intelligence by way of recommending a subsequent step, flagging the deal for evaluate, or triggering a follow-up.
The information layer offers an agent uncooked subject matter. The intelligence layer will give it a head get started, one thing no standalone type, and no platform with no community of this scale, can mirror.
How we take into accounts our platform
There’s so much converting within the business presently. Some platforms will reply by way of ultimate down, establishing walled gardens, proscribing get admission to, and making it tougher for purchasers to get pleasure from AI. We expect the instant requires the other. When brokers can get admission to knowledge, act on behalf of shoppers, and run trade processes, openness and believe topic greater than ever.
Buyer worth above all. We consider consumers will have to have the liberty to select the most productive brokers, integrations, and companions to lend a hand them develop. We’ll all the time put money into world-class first-party brokers from HubSpot. However the most productive agent for a specialised business or workflow will continuously come from the ecosystem. We welcome that.
Open by way of design. We’re running towards a easy usual: anything else you’ll be able to do inside of HubSpot, you will have to be capable of do thru an API. Our intelligence will have to achieve you anywhere you’re employed, inside of or out of doors of HubSpot, immediately or thru apps and brokers constructed on most sensible folks. That’s why we’re dedicated to giving developers get admission to to the similar foundations we construct on.
Relied on by way of default. We’re treating believe and governance as core infrastructure. When a buyer connects a spouse instrument, spins up an agent, or builds one thing customized, they will have to know precisely what it will probably get admission to and what it’s doing. Brokers that act for your behalf are handiest helpful if you’ll be able to believe them.
Those aren’t simply rules. They’re a planned selection about the type of platform we need to be.
What’s to be had nowadays, and what’s coming
Lately: an open, agent-ready platform. HubSpot is open for brokers now. Our APIs and MCP server are reside. Connectors for Claude, ChatGPT, Gemini, and Copilot are turning in actual worth to consumers. Greater than 2,000 apps run throughout our ecosystem, and new brokers are being constructed on most sensible of the platform each and every week.
Coming subsequent: complete API parity. We’re proceeding to make bigger our public API floor in order that each and every capacity of the platform – each and every workflow, each and every motion, each and every piece of context – is out there to the apps and brokers constructed on most sensible folks. No capacity will have to reside handiest at the back of a UI.
The chance forward
The shift to brokers is already going down in each and every GTM crew making an attempt to determine the place the paintings is going now, and in each and every builder deciding which platforms are price making an investment in. We expect the solution comes right down to context. The most efficient brokers would be the ones that perceive a trade the way in which an ideal marketer, gross sales rep, and CSM does: what’s standard, what’s running, what’s converting, what’s labored for firms find it irresistible.
That’s what HubSpot has spent 20 years growing throughout 280,000+ companies. And that intelligence is what we’re opening up – to each and every agent, each and every spouse, and each and every buyer shaping what comes subsequent. We gained’t construct each and every solution. However we’ll construct what each and every solution wishes.
![]()
