The words that survive the room, for a developer-platform decision

If you have five minutes

Whoever signs off on a new connector, a webhook, or an agent with tool access into the portal is not evaluating the feature.
They're asking one question in three different costumes: what happens if this breaks, what happens if this is compromised, and what happens if we can't get rid of it later.

1 of 4

The person who approves this isn't asking what the API does

Whoever signs off on a new connector, a webhook, or an agent with tool access into the portal is not evaluating the feature. They're asking one question in three different costumes: what happens if this breaks, what happens if this is compromised, and what happens if we can't get rid of it later. Answer those three, in that order, or the request sits in an inbox.

2 of 4

Say this, not that

Say: this can now tell you it's ready before it acts.: The Payments Account Public READ API exists so an integration can check whether an account is actually set up to receive payments before it tries to act on that assumption. Casey Hawkins and Chris Carolan's take: "Anything that lets an integration verify state instead of assuming it removes a class of silent failure, which is the kind of failure that surfaces weeks later as a support ticket nobody can reproduce." That is the sentence to bring into the room, because it's the difference between an error a person can act on and one that shows up as an angry customer three weeks later.
Do not say: we'll just automate it and it'll be fine.: The two biggest takes this door leans on both exist because something used to fail silently. Merged records used to change ID without telling anyone. An account could be unready for payments and an integration would find out by failing. Promising automation without naming what it's automating past sounds like a benefit and lands as removing the one checkpoint where a person would have caught the problem.
Ask for the reversible thing first.: Reading a webhook, checking a state, pulling account info — all of that is a read or a notification, not a write. Landing an agent's first tool-use as something reversible (a webhook that notifies, not one that mutates) is what earns the second ask, the one where the agent gets to act.
Name what "beta" actually means here, before someone else does.: Seven of the twenty updates this pillar tracks are still in public or private beta right now. Saying so first, and naming which seven, is what makes the rest of the pitch credible.

3 of 4

What's actually live, and what's still a beta

Update	Status	What it's for
Rendering Locale Dataset Upgrade: JRE to CLDR	Live	Automatic date/number/currency formatting migration; spot-check only if you depend on an exact formatted string
Payments Account Public READ API	Live	Lets an integration verify a payments account is ready before acting
193+ new and 13+ featured apps in the HubSpot Marketplace	Live	The quarter's app refresh, including HubSpot's own agent for Microsoft 365 Copilot
Improvements to Marketplace Submission Eligibility	Live	First-listing bar now measured by active installs, not total installs
Connected Apps: Redesigned App Overview & App Owner Management	Live	Who owns an installed app, and what it can see
HubSpot Connector for Gemini	Live	HubSpot's own connector, shared with two other Hubs
Updates to HubSpot connector for ChatGPT	Live	Same category, shared with three other Hubs
Updates to HubSpot Connector for Claude & Remote MCP Server	Live	Same category, shared with two other Hubs
Updates to HubSpot connector for Claude & remote MCP server	Live	Same category, filed under this Hub alone
Account Name and Creation Date Available from the Account Info API	Live	Two new read-only fields
Location as Restriction: Make Card Location Optional	Live	A card-configuration constraint made optional
FileUpload and FileViewer for UI Extensions	Public beta	Lets a custom card hold files inside the record experience
Public API for CRM Pipeline Rules	Public beta	Shared with three other Hubs
Remote MCP Connectivity for HubSpot Agents	Public beta	How an outside AI product reaches a live portal
Contracts API	Public beta	Shared with two other Hubs
Primary ID Preservation for Merged Records	Public beta	Stable record IDs across a merge
Use Webhooks in custom agents	Private beta	An agent can now send a webhook as one of its tools
Describe an app, deploy it to your portal	Private beta	Describe a page, deploy a real app that reads live CRM data
Pipelines API v1	Upcoming sunset	The path that's closing, not opening
Legacy Private App Creation	Upcoming sunset	Same

4 of 4

Five objections, and the honest answers

"Won't this get expensive once it's running unattended?": Yes, potentially, and that's worth saying out loud before someone finds it on a bill. Use Webhooks in custom agents shipped the same week custom agents started consuming credits against every run. Chris Carolan's take: "A capability expansion and a pricing change landing in the same week is not a coincidence." The honest framing: expanding what an agent can reach and starting to meter every time it runs are the same decision, made twice.
"How do we know an app in the marketplace won't damage the account?": You don't, fully, and neither does HubSpot's own new eligibility bar promise that — it only raises the floor for who gets to list. Chris's take on the warning a customer sees before installing an uncertified app: "Scares the crap out of most people the first time you see it." That warning is the honest signal, not a false alarm to explain away.
"Do we actually need to build this ourselves?": Sometimes no. Casey Hawkins's read on the marketplace refresh, from client work the same week: two sales teams had leads scoring with nobody sure what to do about it, and the entire fix was asking Breeze Assistant what the person had done recently and drafting from there — no new integration built. Chris's structural read on the same update: "But stuff like this is probably going to make it so a lot of integration building just starts to melt away." Ask what's already solved before scoping a build.
"Is this the kind of change that breaks a downstream integration silently?": It has happened before, on this exact platform, which is the honest answer. A merge silently changed a record's ID for as long as anyone can tell, and it stayed invisible because the old ID kept redirecting to the survivor. Chris found out from an AI tool telling him as an aside: "It wasn't until just really like less than a month ago that I did a merge using Claude through the MCP and Claude actually was like, just a heads up. Now these records have different IDs. And I was like, I'm sorry, what?" Primary ID Preservation for Merged Records exists to close exactly that gap, and it's still a public beta.
"What's the actual cost of doing nothing here?": The take on that same ID-preservation update says it plainly: "Silent ID churn is the reason teams decided a HubSpot integration was too hard to maintain and never built one." The cost of not building is the integration that never gets built at all, quietly, because the last one broke and nobody trusted the platform enough to try again.

The sentence to hold on to

Every update on this list either gives you more reach into a live portal or gives you a way to check that reach is safe before you use it. If you can't say which one you're asking for, that's the question to answer before the room asks it for you.

"Anything that lets an integration verify state instead of assuming it removes a class of silent failure, which is the kind of failure that surfaces weeks later as a support ticket nobody can reproduce." — Casey Hawkins + Chris Carolan, Payments Account Public READ API.

"But stuff like this is probably going to make it so a lot of integration building just starts to melt away." — Casey Hawkins + Chris Carolan, 193+ new and 13+ featured updated apps in the HubSpot Marketplace.

"Scares the crap out of most people the first time you see it." — Chris Carolan, Improvements to Marketplace Submission Eligibility.

Where to next

Where this comes from

The show's own update pages. every quote here is the take published at anotherorangemorning.com/hubspot-updates/{slug}, curled live and checked line by line on 2026-09-05
The live HubSpot Updates read, 2026-09-05. every status and first-covered date is the value on the update's own record at the time of that read, calibrated against a known-nonzero control before it was trusted
Chris's steer on this pillar, the nine-Hub walk, 2026-09-05. "Shouldn't this Hub focus on API related updates?" — read as the developer surface, with the admin, account and reporting-UI records filed under the same value treated as tagging debt rather than members

What we do not know yet

Seven of the twenty updates this pillar tracks are still in public or private beta right now. Saying so first, and naming which seven, is what makes the rest of the pitch credible.
"How do we know an app in the marketplace won't damage the account?" You don't, fully, and neither does HubSpot's own new eligibility bar promise that — it only raises the floor for who gets to list.
A merge silently changed a record's ID for as long as anyone can tell, and it stayed invisible because the old ID kept redirecting to the survivor. Primary ID Preservation for Merged Records exists to close exactly that gap, and it's still a public beta.