HubSpot
Also known as: Breeze AI, HubSpot Agent Hub, Breeze Agents, Breeze Studio
AI agent layer across HubSpot's CRM, where ready-made marketing, sales and service agents run on live customer data from day one and custom agents are built by describing them in plain language.
HubSpot is a customer platform spanning marketing, sales, customer service, content, data and revenue, built on a single Smart CRM record. Agent Hub, formerly Breeze Agents, is where its AI agents are turned on and managed: content, prospecting, customer, data, campaign, nurture and revenue agents ship ready-made, with company research, customer health, deal loss, RFP, cross-sell and other agents available from the HubSpot Marketplace.
Because they run on the CRM itself, agents have customer context from the first day with no data connection to set up, and they draw on knowledge vaults holding internal documentation, product details and process guidelines. Teams that need something different build it in agent builder, formerly Breeze Studio, by describing the agent to Breeze Assistant in plain language and refining the proposed goal, actions and knowledge sources on a single canvas, then testing it before publishing.
Oversight is built into the shipped agents rather than left to policy: CRM updates and follow-up drafts are approved after meetings, knowledge base articles are reviewed and approved before publication, the customer agent escalates to a person when needed, and an agent can never act beyond the permissions of the user running it.
HubSpot publishes SOC 2 Type II and SOC 3 through its Trust Center, supports SAML single sign-on with SCIM provisioning, and hosts accounts in the United States, EU, Canada or Australia with paid accounts able to migrate between regions. The developer platform is extensive, with public APIs, client libraries, webhooks, UI extensions and both a hosted CRM MCP server and a local developer MCP server.
Pricing starts free for up to two users, with Starter from $7 per seat per month on an annual commitment and Professional and Enterprise editions carrying seat bundles; agents consume HubSpot Credits, priced at $9 per 1,000 annually and metered per unit of work delivered.
Vendor details
Canonical URL
https://hubspot.com
Category
GTM / revenue agent
Subcategory
Marketing — AI agent layer across CRM
Funding status
Public company, listed on the New York Stock Exchange as HUBS. Reports 306,000 customers worldwide.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Agents and assistants reach email, calendar, shared documents and team conversations in Google Workspace, Microsoft 365 and Slack alongside the Smart CRM. Named connections include LinkedIn Sales Navigator and CRM Sync, WhatsApp, Facebook Messenger, Google Search Console, ad accounts for conversion event syncing, connected social accounts and HubSpot-provided phone numbers, with a Custom Channels API for channels HubSpot does not ship and the HubSpot Marketplace as the general route for third-party apps. The developer platform adds REST APIs generated from published OpenAPI specifications, maintained client libraries, OAuth public apps, private app tokens, a Webhooks Journal API, UI extensions with serverless functions and a CLI. Two MCP servers ship: a hosted remote server at mcp.hubspot.com with read and write access to CRM objects over OAuth, and a local developer MCP server for building HubSpot apps from an AI code editor.
In practice
Your marketing, sales, and service AI come from three different vendors that don't share data. HubSpot runs its agents across one CRM spanning all three.
You want an agent that fits your process, not a generic one. Describe it to Breeze Assistant in plain language, refine the proposed goal and actions, and test it before publishing.
You need AI that can't quietly go too far. Agents approve CRM updates and follow-up drafts with a person, escalate when they should, and never act beyond the permissions of the user running them.
Sources & related URLs
Related / legacy domains
Research sources
Agentic Index coverage score
11.0 / 14 capabilities · 79%
| Integrations & Tool Calling | Full |
|---|---|
|
Breadth across several classes, named on first-party pages. Beyond the CRM itself, agents and assistants reach emails, calendar, shared documents and team conversations in Google Workspace, Microsoft 365 and Slack; the pricing feature inventory names LinkedIn Sales Navigator and CRM Sync, WhatsApp, Facebook Messenger, Google Search Console, ad conversion events synced to ad accounts, connected social accounts, HubSpot-provided telephone numbers and a Custom Channels API for building further channels. The HubSpot Marketplace is the general route, published as the place to connect third-party apps. That spans productivity, messaging, advertising, social, telephony and sales tooling rather than one ecosystem class. Pre-built agents ship with default tools that cannot be edited or rearranged, a real limit on how one agent is configured rather than on how broadly the platform integrates. No first party count of marketplace apps is published. Sourcehubspot.com/products/artificial-intelligence/breeze-ai-assistant, hubspot.com/pricing/suite and ecosystem.hubspot.com/marketplace/appsread 2026-09-02 |
|
| Workflow Orchestration | Full |
|
Multi-step and multi-agent execution on a single canvas. The agent builder, formerly Breeze studio, builds custom agents and agentic automation on one canvas covering everything from simple triggers to complex AI tasks, and HubSpot defines agents as components that follow defined steps to complete structured tasks using the customer's data and tools. Agents hand work to each other across the lifecycle: campaign agent plans a campaign, content agent creates it and nurture agent converts it, while knowledge base agent is documented as requiring an active customer agent subscription because the two work together. Underneath sits the workflow engine, up to 300 workflows on Professional and 1,000 on Enterprise with omni-channel automation, sequences triggerable from workflows and workflow health monitoring. Sourcehubspot.com/products/artificial-intelligence, knowledge.hubspot.com/ai/understand-breeze and hubspot.com/pricing/suiteread 2026-09-02 |
|
| Knowledge Grounding & RAG | Full |
|
Grounding on the customer's own material through two maintained structures. Agents run on the Smart CRM record itself, so output is grounded in live customer data from day one with no setup to connect it, and the data agent answers custom questions about any contact or company by pulling from CRM data, calls, emails and documents. Alongside that sit knowledge vaults, which store reference information agents and custom assistants draw on, explicitly including internal documentation, product details and process guidelines. Customer agent additionally searches the knowledge base, up to 10,000 articles on the top tier, and past contact and contract history before answering. Both are maintained stores rather than per-request assembly. Sourceknowledge.hubspot.com/ai/understand-breeze, hubspot.com/products/artificial-intelligence and hubspot.com/pricing/suiteread 2026-09-02 |
|
| Human Oversight & Guardrails | Full |
|
Shipped approval gates at several points, not a claim of control. Prospecting agent output is approved rather than sent: the customer approves automatic CRM updates and follow-up drafts after every meeting. Knowledge base agent output is reviewed, edited and approved before publication. Customer agent escalates to a person when needed. The pricing feature inventory names email approvals, social media approvals and pipeline approvals for deals as tier-gated features. Underneath these sit two structural limits: admins manage which data AI features can access and can disable that access at account level, and an agent cannot be run by a user lacking permission on the features it depends on, so an agent cannot exceed the authority of the person running it. The marketing line that you decide which channels an agent covers and are always in control is posture; the shipped gates above are the mechanism. Sourcehubspot.com/products/artificial-intelligence, hubspot.com/products/service/ai-knowledge-base-agent and hubspot.com/pricing/suiteread 2026-09-02 |
|
| Security, Identity & Governance | Full |
|
Attestations and access controls are both documented on the vendor's own security page. The Trust Center provides SOC 2 Type II and SOC 3, with no NDA required for SOC 3, backed by annual third-party audits, penetration testing, a formal vulnerability management process and a Bugcrowd disclosure program. Controls: SAML-based single sign-on with Okta, Azure AD and OneLogin named, SSO enforceable for all users with configurable exemptions, SCIM provisioning through Okta covering onboarding and deprovisioning, a permissions model spanning users, teams, CRM objects, properties and security settings, a Security Center for 2FA enforcement, inactive-user audits and partner access control, and Sensitive Data properties with application-layer encryption and field-level access controls. Permission inheritance is documented at the agent layer too: an agent cannot be run by a user who lacks access to the features its configuration depends on. Sourcehubspot.com/security-and-compliance and knowledge.hubspot.com/ai-tools/customize-breeze-agentsread 2026-09-02 |
|
| Observability & Auditability | Full |
|
Three separate records rather than one dashboard. Agent operation: Agent Hub tracks what every agent is doing and delivering in a single view with live status and real outcomes per agent, and the Breeze dashboard reports how AI features are used across the account. Account activity: audit logging is part of the permissions model and covers sensitive-data access. Method: HubSpot publishes model cards documenting how each AI feature works, what data it uses and what safeguards apply, which is disclosure of why rather than only what, and the AI Trust Center hosts them. No per-run trace of an agent's reasoning steps is documented. Sourcehubspot.com/products/artificial-intelligence, knowledge.hubspot.com/ai/use-breeze and hubspot.com/security-and-complianceread 2026-09-02 |
|
| Memory & State Persistence | Partial |
|
Breeze Assistant keeps per user memories (name, job title, tone, writing style, recurring instructions) that the user adds or removes by instruction and can view and delete one by one under Memories, and Breeze projects store chats, shared instructions and reference files, with the assistant automatically storing relevant information as context for future chats and access shared with chosen users. No lifetime or retention is stated for either. Sourceknowledge.hubspot.com/ai/use-breeze-assistantread 2026-09-30 |
|
| Deployment & Data Residency | Full |
|
Genuine customer choice of location across four regions. HubSpot states that accounts are hosted in regional data centers in the United States, the EU, Canada and Australia, and that paid accounts can migrate to a preferred region, with documentation covering how regional hosting works and the limited cases where processing occurs outside the hosting location. That is region selection, a choice rather than legal cover: the Data Processing Agreement with its SCCs and Data Privacy Framework coverage sits alongside it rather than standing in for it. Hosting is multi tenant cloud only, with no VPC, on-premises or customer-tenant deployment, so the choice is of region rather than of infrastructure. Sourcehubspot.com/security-and-compliance and legal.hubspot.com/securityread 2026-09-02 |
|
| Prebuilt Agents, Templates & Packs | Full |
|
A browsable library of named agents aimed at named jobs. The HubSpot Marketplace is published as the place to discover, browse and install agents, and Agent Hub presents every agent in one view where a customer can turn on what has not been activated in a click. The named set is long and job-shaped: content, prospecting, customer, data, campaign, nurture and revenue agents generally available, with ABM landing page, blog research, closing, company research, cross-sell and upsell, customer health, deal loss, request for proposal and sales-to-marketing feedback agents documented in the knowledge base, several in beta. Featured agents are presented as recommendations a customer can download and modify. Prebuilt agents ship with default inputs, instructions and tools that cannot be edited, deleted or rearranged, so adoption is switch-on rather than fork, and the knowledge base notes that certain agents can no longer be installed going forward without listing which. Sourcehubspot.com/products/artificial-intelligence, knowledge.hubspot.com/ai/understand-breeze and knowledge.hubspot.com/ai/use-breezeread 2026-09-02 |
|
| Triggers & Channel Coverage | Full |
|
Wide channel coverage and several genuine intake routes. Channels named on the pricing inventory include live chat, conversational bots, Facebook Messenger, WhatsApp, one-to-one and marketing email, calling with HubSpot-provided numbers and interactive voice response, social posting across connected accounts, and a Custom Channels API for building channels HubSpot does not ship. Work reaches agents automatically rather than only on request: prospecting agent watches accounts for buying signals and reaches out when one appears, sales automation triggers tasks and notifications when deals change stage, sequences can be triggered from workflows, and custom events with up to 500 definitions feed the automation layer on Enterprise. Sourcehubspot.com/pricing/suite and hubspot.com/products/artificial-intelligenceread 2026-09-02 |
|
| Model Flexibility & Routing | Not documented |
|
No customer choice of model, and the providers are not named. HubSpot refers throughout to third-party AI providers in the abstract, stating that they are contractually required to meet its security and privacy standards, process data only to deliver the requested feature and cannot train their models on customer data, and that admins can control what those providers may access and can opt out of HubSpot's own model improvement. No model selector, per agent model choice, bring your own key option or routing policy is offered. Control is over data access rather than over which model runs. HubSpot publishes model cards for each AI feature in its AI Trust Center. Sourcehubspot.com/security-and-compliance and hubspot.com/products/artificial-intelligenceread 2026-09-02 |
|
| APIs, SDKs & MCP Extensibility | Full |
|
Callable from outside through a full developer platform at developers.hubspot.com. REST APIs generated from OpenAPI specifications with the spec itself served at a public endpoint, maintained client libraries including Node, public apps authenticating by OAuth and private apps by access token, a Webhooks Journal API with batched reads and CRM object filtering, date-based versioned APIs, UI extensions with serverless functions, a CLI, and an app marketplace. Two MCP servers ship alongside: a hosted remote server at mcp.hubspot.com giving read and write access to CRM objects over OAuth with PKCE, managed through MCP auth apps in the developer platform, and a local developer MCP server for building apps and CMS assets from an AI code editor. Product-side surfaces include a Custom Channels API and a marketing email single send API named on the pricing feature list. The fact that prebuilt agents' tools cannot be edited concerns how one agent is configured, not whether the platform is callable. Sourcedevelopers.hubspot.com/ai-tools, developers.hubspot.com/docs/api/client-libraries and developers.hubspot.com/spotlightread 2026-10-01 |
|
| Testing, Debugging & Optimization | Partial |
|
A quality gate before an agent goes live, with no readable comparable result. The agent builder documentation states that after a configuration is generated a customer adjusts instructions, adds actions and tests the agent before publishing, which is a real pre-release check the customer runs on the agent itself rather than a claim about internal testing. Around it sit A/B testing on marketing email, custom reporting, workflow health monitoring on the higher tiers, and the per-agent performance view in Agent Hub for keeping what works. Nothing surfaces a score, pass rate, judge verdict or other readable comparable result from that test, and no sandbox, replay, regression suite or evaluation harness over agent behavior is documented. Sourcehubspot.com/ai/create-agents-with-breeze-assistant and hubspot.com/pricing/suite; knowledge.hubspot.com/ai/create-agents-with-breeze-assistantread 2026-09-02 |
|
| Browser & Computer Use | Not documented |
|
Every documented action runs through a programmatic route. Agents act on the Smart CRM and reach third-party systems through integrations, the marketplace and documented APIs, and the extensibility surface is an API and MCP one throughout; the developer MCP server drives an AI code editor rather than a browser. No agent is described operating software through its interface, controlling a browser, or navigating a third-party console where no programmatic route exists. Sourcehubspot.com/products/artificial-intelligence and developers.hubspot.comread 2026-09-02 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Recent platform changes
Agent Builder custom agents, the Prospecting Agent and the Data Agent are now part of HubSpot's refreshed Starter plan. Prospecting and research runs draw on HubSpot Credits.
Bears on: Agent capability
View sourceHubSpot released its August 2026 product updates, which include new features for its Customer Agent and Breeze AI. The update introduces the ability to add custom context to Breeze AI via free-form text or file uploads. The HubSpot Connector for Claude and Remote MCP Server now supports querying and creating core CRM records such as Contacts, Companies, Deals, Tickets, Notes, Tasks, and Meetings.
Bears on: MCP / tool calling / API
View sourceHubSpot expanded its Claude connector for the remote MCP server, adding Partner Client read access for partners and solutions team members. The update also introduces verified domain restrictions specifically for Claude Enterprise accounts.
Bears on: MCP / tool calling / API
View sourcePricing
Free for up to 2 users. Starter from $7 per seat per month on an annual commitment, $20 monthly. Professional from $1,300 per month including 6 seats; Enterprise from $4,700 per month including 8 seats.
seats plus HubSpot Credits
Included quota
Included HubSpot Credits per month by edition, as published on the pricing page: 500 on Starter, 5,000 on Professional, 10,000 on Enterprise. Corrected this pass from 500 / 3,000 / 5,000. Agent Hub itself is included in every paid edition; Breeze Assistant is available on every plan including free. Availability by agent is now split rather than uniformly Professional-plus: prospecting agent and data agent are marked Starter and above, customer agent and Data Studio Professional and above.
What is public
A complete, retrievable rate card. The pricing page publishes four editions with prices, seat bundles and additional seat costs, the included credit allowance for each, the credit rate of $9.00 per 1,000 paid annually, and a per-feature consumption table naming credit costs for the customer, data and prospecting agents and for Data Studio, with a credit estimator keyed to CRM contact volume. Professional and Enterprise show a Talk to Sales route while Free and Starter carry a direct buy path, which is why the model reads self-serve at entry with sales-led upper tiers.
Billing mechanics
Two axes. Subscription: free for up to 2 users, Starter from $7 per seat per month annually or $20 monthly, Professional from $1,300 per month with 6 seats included and additional core seats from $45, Enterprise from $4,700 per month with 8 seats included and additional core seats from $75. Consumption: agents and many AI features draw on a shared HubSpot Credits pool at $9.00 per 1,000 paid annually, metered per unit of work delivered. Published consumption rates are 50 credits per conversation the customer agent resolves, 100 credits per recommended outreach for one lead from the prospecting agent, 10 credits per smart properties run for the data agent, and 25 to 200 credits per dataset action run in a workflow, segment, CRM sync or export for Data Studio. At the annual credit rate that is about $0.45 per resolved conversation and $0.90 per recommended lead. The outcome-based framing introduced in April 2026 has been absorbed into the credit pool rather than sitting alongside it.
Cost watchouts
Credit consumption is the real variable: 50 credits per conversation the customer agent resolves, 100 per recommended outreach for one lead, 10 per smart properties run and 25 to 200 per dataset action, all drawn from one pool at $9 per 1,000 annually. Roughly 500 resolved conversations a month is about 25,000 credits, well beyond the 5,000 included on Professional. Seats are a second axis: Professional and Enterprise bundle 6 and 8 seats, with additional core seats from $45 and $75. The customer agent requires Professional or above.
Variable cost rationale
Spend at the agent layer rises directly with resolved conversations, recommended leads and dataset actions, all drawn from one credit pool on top of the seat subscription, and the metered units are defined by the vendor. The subscription floor is low and public, so almost all of the uncertainty sits in consumption.
Additional watchouts
Cost is driven by credit consumption rather than by the subscription, and the metered units are defined by HubSpot: what counts as a resolved conversation or a recommended lead is the vendor's definition. The customer agent requires Professional or above, so the entry price understates the cost of running the flagship agent.
Overage / add-ons
Once the included credit allowance is used, further credits are bought in capacity packs up front or as consumed with pay-as-you-go. Published rate is $9.00 per 1,000 credits paid annually. Recorded but not re-confirmed this pass: an earlier first-party company-news post described a 72-hour reopen rule under which a customer replying more than 72 hours after the agent's last message reopens the thread and bills a further resolution.
Sales call required
Mixed (some tiers require a call)
Free / trial
Free tier for up to 2 users with no credit card required, plus a free 28-day trial of the Breeze agents on Professional and Enterprise. Breeze Assistant is available on every plan including the free CRM.
Lowest paid plan
Starter, from $7 per seat per month on an annual commitment ($20 monthly), including 500 HubSpot Credits.
Commercial notes
HubSpot reports 306,000 customers worldwide. Published agent outcomes on the pricing and product pages: over 70% of conversations resolved automatically by the customer agent and 39% faster ticket resolution, up to 2x higher response rates for the prospecting agent, and 65% more sales leads per month. Founded 2006, Cambridge, Massachusetts; public on the NYSE as HUBS. HubSpot followed Sierra, Intercom and Zendesk into outcome-based AI pricing in April 2026, and has since folded that metering into its credit pool.
Key ambiguities
Total cost depends heavily on volume, which Hub and tier you are on, and how a resolved conversation or a recommended lead is defined. Agents and many AI features draw on one HubSpot Credits pool, which is separate from HubSpot's seat and contact subscriptions.
Cancellation / refund
Outcome pricing applies to all customers (no opt-out/grandfathering); free 28-day Breeze agent trial for Pro/Enterprise; credits billed monthly ($10/1,000) or ~10% cheaper annually ($9/1,000); underlying Hub subscriptions carry their own terms
Support SLA / resale
Agents available on Professional/Enterprise Hubs; Breeze is embedded in HubSpot Smart CRM (customer history, deals, tickets, marketing engagement) for context; HubSpot connectors include Anthropic's Claude and ChatGPT deep research; 248,000+ customers in 135+ countries
Missing data
The 72-hour resolution reopen rule and the earlier per-outcome dollar framing come from an April 2026 company news post and are not confirmed on the live pricing page. Marketplace app counts and the full agent availability matrix by tier are not stated.
Related vendors
- 11x — AI digital workers (Alice for outbound SDR work, Julian for inbound…
- 1up — AI answer engine for sales teams that automates RFPs and security…
- Actively AI — GTM superintelligence platform that trains a custom reasoning model…
- Aircover.ai — AI sales assistant that delivers real time in call coaching, a…
- AiSDR — AI SDR platform with multichannel outreach automation, quota-visible…
- Akkari — Autonomous customer ops agent that captures every commitment,…
Alternatives to HubSpot
The closest documented capability profiles to HubSpot among GTM and revenue agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Gainsight10.5 / 14A lighter documented profile than HubSpot
- Amplemarket10.0 / 14A lighter documented profile than HubSpot
- Creatio11.0 / 14Adds documented Model Flexibility & Routing
- Fireflies.ai10.0 / 14A lighter documented profile than HubSpot
- Hightouch11.0 / 14Fuller documented coverage on Testing, Debugging & Optimization
- Optimizely12.0 / 14Fuller documented coverage on Memory & State Persistence and Testing, Debugging & Optimization
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded