Skip to main content

​
How it works

​
What is HubSpot Agent Hub?

HubSpot Agent Hub is HubSpot’s platform for AI agents: pre-built agents, custom agents you build yourself, and agentic workflows, managed in one place across the customer journey. It sits inside the CRM your team already runs on, which is what makes it interesting for a partner program: the agent starts from live CRM data rather than from an export. Three pieces matter for partner plays:
  • Pre-built agents cover the direct motion out of the box: AEO, Data Agent, Prospecting Agent, Deal Progression, and Customer Agent.
  • Custom agents are built in the agent builder from four ingredients: instructions (the agent’s role, goal, and output), actions (what it is allowed to do), knowledge (brand, documents, knowledge vaults, CRM context), and inputs (the per-run data such as a deal name or close date).
  • Agentic workflows put an agent inside HubSpot automation. You configure triggers and actions the way you already do, and add an agent as a step, so a play runs on a CRM event or a schedule instead of when someone remembers.
Agent Hub is in beta on Professional and Enterprise tiers, and agent runs consume HubSpot Credits. Check HubSpot’s Agent Hub documentation for the current requirements.

​
Where Introw fits: the MCP server inside the app

An agent’s actions decide what it can actually do. HubSpot supports three sources: HubSpot’s own actions, defaults like reading CRM records or browsing the web, and MCP connectors, which link an agent to an external system over the Model Context Protocol. The Introw HubSpot app ships an MCP server as part of the app. Install or update the app, and Introw MCP appears in the Connectors tab when you add an action to an agent. Introw is the first partner-management app to make its data and actions available to Agent Hub this way, so partner plays stop being something you bolt on next to HubSpot and become something HubSpot agents run natively. That single connection is what closes the loop:
  • HubSpot reads its own data natively. The agent runs inside HubSpot, under the permissions of the person running it, so nothing about this widens Introw’s OAuth scopes.
  • Introw is where the agent acts. Through the MCP server the agent can search partners, tiers, goals, commissions and marketing funds, look up partner tasks and timeline activity, read and process form submissions such as deal registrations and shared leads, post partner-facing comments, create and update tasks, update partner fields and CRM object properties, submit any partner form, and pull the right enablement content.
  • The CRM stays the system of record. Every write lands back where your reporting already looks.
See the MCP overview for the server itself, and Connect HubSpot for the app install.

​
The leverage: what only lives on the direct side

A partner platform can only act on what partners give it. The reason Agent Hub is worth building on is the data sitting on the other side of the business, which the agent reads natively:

​
The plays

Fifteen workflows worth building, grouped by where they sit in the partner motion. Each one names its HubSpot trigger, what it reads on the direct side, and what it writes back through Introw.

​
Bringing partners into the motion

Three plays that get a partner and an opportunity onto each other, in both directions: a rep who needs a partner, a partner who needs a rep, and an account signal that should wake a partner up.
Trigger: a direct deal is created or reaches qualification, or the rep asks.Reads: products and amount on the deal, account industry, size and geography, what the rep heard the customer needs, competitor in play.The agent scores every partner against this specific deal: who holds a live certification on the products on it, who has closed deals like it in this segment and size band, who covers the geography, who works this vertical, who is trending up rather than coasting on old numbers, and who has room this quarter. It returns a ranked shortlist of two or three with the evidence behind each, writes it onto the deal record, and drafts the introduction.Recency is weighted over volume, so a partner who won three deals like this in the last two quarters outranks one who won twelve of them three years ago. In a young program with thin attribution history, certification, geography and category carry the ranking until there are enough wins to read a pattern from.Why it matters: this is the only play whose main user is the direct rep. Co-sell attach is a push metric today, with the partner team chasing sellers to involve partners, and this makes it a pull. It also spreads opportunity on evidence rather than on whichever partner the rep happens to remember, which is the quiet mechanism behind partner concentration.
Trigger: a partner shares a lead.Reads: owner territory and current load, win rate by segment and rep, open pipeline per AE.The agent qualifies the lead, then picks the AE who is genuinely right for it on territory, segment win rate, and who has room this week. It creates the task and tells the partner who owns it and by when they will hear back.The commitment back to the partner is the point. A named owner and a stated response time turn a lead submission from a hopeful gesture into a transaction with terms, and the agent chases the rep if the terms slip.Why it matters: a shared lead that sits unworked for a week teaches the partner not to share the next one. Rep capacity exists only in the CRM, which makes this the one routing decision a partner platform cannot make on its own.
Trigger: a buyer intent signal fires on an account a partner owns or overlaps.Reads: funding, M&A, exec change, expansion, hiring and tech investment, visitor intent on your site.The partner’s account just raised a round, moved a CRO, or started hiring against the problem you solve. The agent tells the partner what happened, why it matters for them specifically, which play to run, attaches the asset, and creates the task.It only fires where the partner has standing: an existing customer, an open registration, or a mapped overlap. A partner who gets alerted about an account they have no relationship with learns to ignore the next one, so the bar for firing is deliberately high.Why it matters: partners get the same intent-driven prompts the direct team gets. Nothing in the partner stack has this data, and it is the cheapest source of net-new partner pipeline you can put in their hands.

​
On the live deal

Five plays for the window where the deal is in flight and both sides need to stay sharp: the partner coached off real wins, the rep fed the partner’s questions, and neither of them guessing what the other just did.
Trigger: the stage sits unchanged, or the close date passes, on a partner deal.Reads: closed-won deals in the same segment, competitor and loss reasons, winning line items and discount, the playbook that rep ran.The agent finds the deals your direct team already won that look like this one, extracts what moved them at this exact stage, and hands the partner two or three specific next moves plus a drafted message to the buyer. Not generic stage advice: the move that worked on the last four deals like theirs.The same pass reads the losses. If deals in this segment reliably die on a security review or a procurement step, the coach raises it before the partner walks into it, with the asset that cleared it last time.Why it matters: direct reps get coached weekly off real win data. Partner sellers get a portal full of PDFs, and the win-rate gap shows it. This is the difference between advice and evidence, delivered on the deal rather than buried in a training module nobody opens. See deal coaching for the wider pattern.
Trigger: a direct rep’s call or meeting ends on a partner-sourced deal.Reads: call transcript and recap, deal stage and next step, the rep’s notes.The agent reads what the customer actually said on the demo, writes a partner-safe summary, and pushes it to the partner’s portal timeline with a task for anything the partner needs to do. The rep writes nothing and changes none of their habits.The filter matters as much as the summary. Internal pricing debate, forecast commentary and competitive positioning stay out. What the customer asked for, what worried them, and the one thing the partner is best placed to influence go in.Why it matters: the partner learns within the hour that the demo happened, how it went, and what is needed from them. Silence after handoff is the single biggest reason referrers and co-sell partners stop bringing deals, and this ends it without adding a step to anyone’s day.
Trigger: a partner comments on a live deal.Reads: deal owner and stage, the rep’s open tasks and load, recent activity on the account.Introw’s partner support agent already answers the routine questions. This play handles what it should not: a live deal question the AE has to own. It classifies the comment, attaches the deal context, assigns it to the right rep with a drafted reply, and puts an SLA on it.Where the rep does not respond it escalates on the timer, to the partner manager first and the sales manager second, so a partner is never left waiting on someone who is on holiday.Why it matters: partner questions enter the direct team’s actual work queue instead of dying in a portal thread nobody owns. Response time on deal-blocking questions is what partners judge a vendor on hardest, and it is usually invisible to the people being judged.
Trigger: the buyer opens, downloads, returns to, or lets a quote expire on a partner deal.Reads: quote view and download events, quote status and expiry, line items and discount applied.The buyer just opened the quote the partner sent. The agent nudges the partner within the hour with the right follow-up, while the interest is live. If the quote goes untouched, or the expiry approaches, it switches to a different message and pulls the rep in.Reading the line items lets it be specific rather than generic. A buyer who keeps returning to a quote without signing is usually stuck on scope or price, so the partner gets their tier’s discount authority alongside the nudge and knows what they can move without asking.Why it matters: quote engagement is one of the strongest buying signals in the CRM, and partners currently have no way to see it. Acting on it in the hour rather than the week is straightforward pipeline velocity, on deals that are already late-stage.
Trigger: a stage change on a referral-sourced deal.Reads: the rep’s notes on why it moved, the next step on the deal.Introw already notifies on stage change. The Agent Hub version adds the part that builds trust: why it moved, what happens next, and what, if anything, the referrer could help with. Written so it protects credit without dragging them into the sales cycle.Tone is calibrated to the audience. A referrer is not a seller, so the update reads like a note from a colleague rather than a pipeline report, and it never asks them to do work they did not sign up for.Why it matters: referrers do not want a status field, they want to know their introduction is being treated well. That feeling decides whether a second introduction ever arrives, and it is almost entirely a communication problem rather than a product one. See deal updates for the built-in version.

​
After the win

Four plays for the moment a partner is most willing to commit to the next thing, and most likely to be dropped instead.
Trigger: a partner-sourced deal closes won.Reads: what the direct team expands into next, line items sold, segment expansion patterns.The agent congratulates with the real commission figure and tier movement, then reads what your direct team typically sells into this segment next and hands the partner the specific follow-on account or product, with the evidence behind the suggestion.It also names the pattern in the partner’s own book. If their last four wins were all mid-market manufacturers, that is the shape to chase again, and the agent lists which of their existing accounts match it.Why it matters: it turns the celebration into the next registration instead of a full stop. The window right after a win is when a partner is most willing to commit to the next one, and almost no program uses it for anything but a congratulations email.
Trigger: a referral closes won.Reads: why the deal actually closed, call transcripts and decision moments, competitor displaced.A short, genuinely celebratory note back to the referrer that does one useful thing: it explains why this one won. Not the amount, the reason. What the customer’s trigger was, which objection nearly stopped it, and what finally clinched it.Then it names the shape to look for next: the company profile, the buying trigger and the persona that made this work, described concretely enough that the referrer recognises it in their own network the same day they read it.Why it matters: referrers guess at what a good introduction looks like, and mostly guess wrong. Telling them exactly why the last one worked is the highest-yield thing a referral program can do, and the moment right after a payout is when they are most receptive to hearing it.
Trigger: a deal closes won with an implementation partner attached.Reads: line items actually sold, sales call transcripts and promises, the rep’s notes and scope.The agent builds the go-live plan on the deal record, scoped from what was genuinely sold and what the rep promised on the calls, checks the assigned consultants hold the right certifications, and creates the scoped tasks on both sides.The promises are the valuable part. Commitments made verbally on a sales call are the usual cause of a delivery blow-up, and they sit in a transcript nobody reads once the deal is closed.Why it matters: implementation partners inherit a forwarded email and a guess. Handing them the sale’s actual content is the difference between a clean go-live and a rescue, and go-live rate is what drives renewal, expansion and the stickiness a deep implementation creates.
Trigger: a renewal date approaching on a partner-sourced customer.Reads: subscription and invoice state, CSM notes and health, support ticket history.The agent briefs the partner on where the account stands, what the CSM is worried about, and what the expansion case looks like, then creates the joint tasks on both sides ahead of the conversation.Where the partner did the implementation it flags the delivery history too, because an unresolved go-live issue is usually the thing standing between a renewal and an expansion, and the partner is the only one who can clear it.Why it matters: partners who sourced a customer usually hear about a renewal risk after it churns. Inverting that turns the partner into a retention asset instead of a party with no visibility and no role at exactly the moment they are most useful.

​
The operating rhythm

Three plays for the partner manager, each one dependent on live amounts, live scoring or live owners to be worth running at all.
Trigger: three chained agentic workflows on meeting booked, meeting completed, and thirty days after.Reads: live pipeline and amounts, the QBR recording and transcript, the direct team’s view of those accounts.Seven days out, the pack builds itself from pipeline, goals, commissions and what the direct team sees on the same accounts, with the risks surfaced at the top rather than buried on slide nine. Prep tasks land on the owner’s list at the same time.When the meeting is marked complete, the agent reads the recording and turns it into partner tasks, a portal comment and a follow-up email the same day. Thirty days later the third workflow checks what actually moved against what was agreed and reopens anything that slipped.Why it matters: two to four hours of assembly per QBR becomes fifteen minutes of editing, which is what makes full partner coverage feasible instead of the usual twenty to thirty percent. The thirty-day check is the step nobody does manually, and it is the one that makes the next QBR a review rather than a reset. See QBRs and meeting prep.
Trigger: a deal closes won, or a monthly schedule.Reads: deal score and health, comparable wins from the direct team, ICP fit on the partner’s open accounts.The agent computes the gap to the next tier, then does the part that makes it actionable: it ranks the partner’s open deals by how likely they are to land in time, using your own deal scoring and how comparable deals actually went for the direct team.Where the open pipeline does not cover the gap it looks the other way, and names accounts in the partner’s book that fit the ICP but have no deal on them yet. The nudge arrives with a route rather than just a number.Why it matters: a tier threshold only motivates when the partner can see which two deals get them there and believes they can win them. Live amounts, deal scoring and comparable outcomes all sit in the CRM, which is exactly why this play belongs on the direct side rather than in a portal widget.
Trigger: a HubSpot owner is deactivated.Reads: live owner records and teams, book size and revenue per manager, territory coverage.The moment a partner manager is deactivated, their book is redistributed by size, territory and current load rather than dumped on whoever sits nearest, and the new manager is introduced on each partner timeline.It carries the context forward too. Open tasks, in-flight registrations and anything promised at the last QBR move with the partner, so the new manager starts informed instead of cold and the partner never has to explain their own history back.Why it matters: partners going quiet after their manager leaves is one of the most predictable and least-managed causes of channel churn. Handling it on the trigger, within the hour, removes the gap entirely rather than shortening it.

​
Build the first one

1

Install or update the Introw HubSpot app

The MCP server ships with the app, so HubSpot only offers it once the app is installed and authorized with current scopes. See Connect HubSpot, and re-authorize through Install update if the connector does not appear.
2

Create the agent in Agent Hub

In HubSpot, go to Agents > Agent Hub > Agents and create an agent, or clone a pre-built one. Write the instructions as the play: the trigger you expect, what to read, what to write back, and what to leave alone.
3

Add Introw as an action

In the agent’s Actions section, click Add action, open the Connectors tab, hover Introw MCP and choose Connect and add, then complete the authorization. Add only the tools the play needs; a narrow action list keeps the agent predictable.
4

Give it knowledge

Point the agent at the context it should not have to be told twice: rules of engagement, tier and discount authority, partner-safe language. Introw’s knowledge base holds the partner-facing version of the same material.
5

Put it on a trigger

Move the play from on-demand to automatic with an agentic workflow: configure the HubSpot trigger, add the agent as a step, and let it run on deal stage, quote engagement, owner deactivation, or a schedule.
6

Cost-model before it goes always on

Agent runs consume HubSpot Credits. Run the play manually on a handful of records first, check what one run costs, and only then enroll the whole pipeline.

​
Governance

The safeguards are the same ones every Introw agent runs under, and they hold when the caller is a HubSpot agent rather than a person in a chat window.
  • The agent inherits a person’s access. The HubSpot side runs under the permissions of the user running it, and the Introw side runs under the OAuth connection’s partner scope. Neither widens.
  • Sensitive actions still wait for a human. Introw’s capability matrix marks each action Read-only, Allowed, Approval required, Write/Delete, or Blocked, so an agent that drafts freely can still be blocked from approving an MDF request. See governance and trust.
  • Everything is auditable. Writes land on the CRM timeline and the partner timeline, which is where your reporting already looks.
  • Partner-facing output is filtered on purpose. Plays that read internal material, transcripts especially, publish a partner-safe summary rather than the source.

​
Key takeaways

Key takeaways

  • Agent Hub is HubSpot’s home for pre-built agents, custom agents, and agentic workflows, built from instructions, actions, knowledge, and inputs.
  • Introw is the first partner-management app with an MCP server for Agent Hub, so a HubSpot agent can act on your partner program through the app you already installed.
  • The leverage is direct-side data: closed-won history, transcripts, playbook logs, buyer intent, rep capacity, quote engagement, subscriptions, and deal scoring, none of which a portal-first PRM can see.
  • Fifteen plays, across bringing partners into the motion, the live deal, after the win, and the operating rhythm.
  • Cost and control: runs consume HubSpot Credits, and every Introw action stays inside the capability matrix and the signed-in user’s scope.

​
Frequently asked questions

Agent Hub is HubSpot’s platform for AI agents. It brings pre-built agents (AEO, Data Agent, Prospecting Agent, Deal Progression, Customer Agent), custom agents you build in the agent builder, and agentic workflows into one place, organized along the customer journey so you can see where AI is active and what it produced. It is in beta on Professional and Enterprise tiers and consumes HubSpot Credits.
Through MCP. The Introw HubSpot app ships an MCP server, so once the app is installed, Introw MCP shows up in the Connectors tab when you add an action to an agent. You connect it once, authorize it, and the agent can then use Introw’s tools as actions.
Read partners, tiers, goals, commissions, marketing funds, partner tasks, and timeline activity; read and process form submissions such as deal registrations and shared leads; post partner-facing comments; create and update tasks; update partner fields and CRM object properties; submit any partner form; and retrieve enablement content. Every one of those runs through the capability matrix, so you decide which execute on their own and which route to a human.
No. Introw’s own AI agents and the MCP server run these motions from Claude, ChatGPT, Slack, Teams, or the portal without Agent Hub. Agent Hub adds one thing the others cannot: native access to direct-side CRM data such as call transcripts, playbook logs, quote engagement, and rep capacity, without anyone widening an integration’s scopes.
No. The Introw connection authenticates as a user over OAuth and inherits that user’s partner scope, exactly like connecting Claude or ChatGPT. Actions that change data are governed by the same capability matrix, and sensitive ones can be set to require approval or be blocked outright.
Partners do not touch your Agent Hub. They get the mirror image through Partner Connect: their own assistant, their own chat tools, and their own CRM, scoped to their own data. Several plays here are designed to land in exactly that surface, which is why the partner sees a portal update or a task rather than a HubSpot notification.
Start with one that has a clean trigger and an obvious owner. Partner recommendation on a direct deal proves the value to the sales team, demo transcript to portal update proves it to partners, and shared-lead routing on live capacity fixes the response-time problem partners complain about most. Leave the scheduled plays until you have cost-modelled a run.

​
Going deeper

HubSpot integration

Two-way sync, cards, workflow actions, and everything else Introw runs inside HubSpot.

MCP overview

The OAuth-protected server the connector talks to, and its full tool surface.

HubSpot workflow actions

The classic-workflow counterpart: Introw actions you drop into a HubSpot workflow.

Skill library

The same plays as named skills you can run from Claude Code today.