HubSpot vs Salesforce (2026)

Beyond the CRM itself, agents and assistants reach emails, calendars, shared documents and team conversations in Google Workspace, Microsoft 365 and Slack. Other named connections include 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 for connecting third-party apps across productivity, messaging, advertising, social, telephony and sales tools, though no count of marketplace apps is published. Prebuilt agents ship with default tools that cannot be edited or rearranged, which limits how one agent is configured, not how broadly the platform integrates.

Sourcehubspot.com/products/artificial-intelligence/breeze-ai-assistant, hubspot.com/pricing/suite and ecosystem.hubspot.com/marketplace/appsread 2026-09-02

Agents act on the customer's systems through actions. Agentforce Builder connects agents to data, channels and existing Salesforce assets including Flows, Apex, MuleSoft APIs and prompt templates, and out of the box actions cover common jobs such as order and reservation management.

Sourcesalesforce.com/agentforce/agent-builderread 2026-09-18

The agent builder, formerly Breeze studio, builds custom agents and agentic automation on one canvas, 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, with the campaign agent planning a campaign, the content agent creating it and the nurture agent converting it, and the knowledge base agent requires an active customer agent subscription because the two work together.

Underneath sits the workflow engine, with up to 300 workflows on Professional and 1,000 on Enterprise, 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

Agent Script defines how work moves through an agent: every request begins at an agent router that sets initial variables and classifies the request, then the agent processes a subagent and transitions between subagents along paths the script defines. Across agents, an orchestrator agent can invoke specialized agents in other Salesforce orgs and outside Salesforce through the Agent API, and a flow action can invoke an agent as one step of a longer automation.

Sourcedeveloper.salesforce.com/docs/ai/agentforce/guide/ascript-flow.mdread 2026-09-18

Channels 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 channels HubSpot does not ship. Work also reaches agents automatically.

The 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, up to 500 definitions on Enterprise, feed the automation layer.

Sourcehubspot.com/pricing/suite and hubspot.com/products/artificial-intelligenceread 2026-09-02

Headless agents are called without a human trigger from a user interface. An Apex trigger or a scheduled job calls the agent as part of a background process, and a standard AI Agent action lets any flow, including one started by a record change, invoke an active agent with a prompt.

Sourcedeveloper.salesforce.com/blogs/2025/04/build-headless-agents-with-the-agent-api.htmlread 2026-09-18

Agents run on the Smart CRM record itself, so output is grounded in live customer data from day one with no setup, and the data agent answers custom questions about any contact or company from CRM data, calls, emails and documents. Knowledge vaults store reference information that agents and custom assistants draw on, including internal documentation, product details and process guidelines. The customer agent also searches the knowledge base, up to 10,000 articles on the top tier, and past contact and contract history before answering.

Sourceknowledge.hubspot.com/ai/understand-breeze, hubspot.com/products/artificial-intelligence and hubspot.com/pricing/suiteread 2026-09-02

Agents are grounded in the customer's own material. An Agentforce Data Library connects an agent to trusted unstructured or semi-structured sources, such as files or a knowledge base, which the customer adds to the library and the agent retrieves from when it answers, and new material is added to the library with no retraining.

Sourcedeveloper.salesforce.com/docs/ai/agentforce/guide/headless-examples-agent-adl.mdread 2026-09-18

Breeze Assistant keeps memories for each user, such as name, job title, tone, writing style and recurring instructions, which the user adds or removes by instruction and can view and delete one by one under Memories. Breeze projects store chats, shared instructions and reference files, the assistant automatically stores relevant information as context for future chats, and access can be shared with chosen users. No lifetime or retention is stated for either.

Sourceknowledge.hubspot.com/ai/use-breeze-assistantread 2026-09-30

Agent Script variables let an agent deterministically remember information across conversation turns and keep context throughout the session, shared by every subagent, and a session ID carries that context across invocations. Data 360 and Customer 360 give agents live customer records, which are the business records themselves, and Salesforce does not describe a memory the agent keeps across sessions.

Sourcedeveloper.salesforce.com/docs/ai/agentforce/guide/ascript-ref-variables.mdread 2026-09-18

Prospecting agent output is approved, not sent automatically, with the customer approving CRM updates and follow-up drafts after every meeting. Knowledge base agent output is reviewed, edited and approved before publication, and the customer agent escalates to a person when needed.

Email approvals, social media approvals and pipeline approvals for deals are tier-gated features. Admins manage which data AI features can access and can turn that access off for the whole account, and a user who lacks permission on the features an agent depends on cannot run it, so an agent never exceeds the authority of the person running it.

Sourcehubspot.com/products/artificial-intelligence, hubspot.com/products/service/ai-knowledge-base-agent and hubspot.com/pricing/suiteread 2026-09-02

Escalation to a person is a built-in agent utility: utils.escalate hands the conversation to a human service rep through an Omni-Channel routing connection the customer defines, and instructions tell the agent when to call it. Agent Script structures what the agent may do, with every request passing an agent router that classifies it before any subagent runs and explicit transitions between subagents, and Salesforce's own guidance recommends a person check model responses before they are shared in most use cases.

Sourcedeveloper.salesforce.com/docs/ai/agentforce/guide/ascript-ref-utils.mdread 2026-09-18

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 include SAML-based single sign-on with Okta, Azure AD and OneLogin, SSO that can be enforced for all users with configurable exemptions, SCIM provisioning through Okta for 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. Permissions carry through to agents, so a user who lacks access to the features an agent depends on cannot run it.

Sourcehubspot.com/security-and-compliance and knowledge.hubspot.com/ai-tools/customize-breeze-agentsread 2026-09-02

Salesforce holds attestations scoped to the agent layer itself, including a SOC 2 report and a SOC 3 report for the Einstein Platform and Agentforce on Hyperforce, an ISO 42001 certificate covering the Salesforce AI Platform and Agentforce, a PCI attestation of compliance for Data Cloud, the Einstein Platform and Agentforce, a C5 report, an IRAP assessment and Spain ENS High for the Einstein Platform and Agentforce, alongside FedRAMP High and Moderate and ISO 27001 categories.

Sourcecompliance.salesforce.com/enread 2026-09-18

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. Audit logging is part of the permissions model and covers sensitive-data access. HubSpot publishes model cards in its AI Trust Center that explain how each AI feature works, what data it uses and what safeguards apply. No per-run trace of an agent's reasoning steps is described.

Sourcehubspot.com/products/artificial-intelligence, knowledge.hubspot.com/ai/use-breeze and hubspot.com/security-and-complianceread 2026-09-02

Every step of an agent interaction is recorded. The Agentforce Session Tracing data model and the Generative AI Audit and Feedback data model capture turns, messages, LLM calls, actions, metric scores and feedback, and the Session Trace OTel API exports a full session as OpenTelemetry-format JSON for any observability tool or OTLP collector. Trace files can also be managed from the command line during development, and the Einstein Trust Layer keeps an audit trail of generations.

Sourcedeveloper.salesforce.com/docs/ai/agentforce/guide/otel-api.mdread 2026-09-18

Accounts are hosted in regional data centers in the United States, the EU, Canada and Australia, and paid accounts can migrate to a preferred region, with guidance on how regional hosting works and the limited cases where processing happens outside the hosting location. The Data Processing Agreement, with SCCs and Data Privacy Framework coverage, sits alongside the region choice. Hosting is multi tenant cloud only, with no VPC, on-premises or customer-tenant deployment.

Sourcehubspot.com/security-and-compliance and legal.hubspot.com/securityread 2026-09-02

Salesforce runs on Hyperforce, its public-cloud infrastructure, where customer data is generally stored in the country where the customer's org is located, provided the services in use are available in that Hyperforce country, and a secure backup of an org can be held in a secondary Hyperforce region.

Sourcesalesforce.com/platform/public-cloud-infrastructureread 2026-09-18

The HubSpot Marketplace is where customers discover, browse and install agents, and Agent Hub shows every agent in one view, where a customer can turn on any not yet activated with a click.

Generally available agents include content, prospecting, customer, data, campaign, nurture and revenue agents, and 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 are also listed, several in beta. Featured agents are recommendations a customer can download and modify.

Prebuilt agents come with default inputs, instructions and tools that cannot be edited, deleted or rearranged, so adopting one means switching it on, not forking it, and certain agents can no longer be installed, though which ones is not listed.

Sourcehubspot.com/products/artificial-intelligence, knowledge.hubspot.com/ai/understand-breeze and knowledge.hubspot.com/ai/use-breezeread 2026-09-02

Agentforce Builder starts agents from out of the box subagents, actions and templates that cover common jobs such as order management, reservation management and general FAQs, which the customer then extends and refines with natural language instructions.

Sourcesalesforce.com/agentforce/agent-builderread 2026-09-18

Customers cannot choose the model, and the providers are not named. HubSpot refers to third-party AI providers in general terms, saying they must contractually meet its security and privacy standards, process data only to deliver the requested feature and cannot train their models on customer data.

Admins can control what those providers may access and can opt out of HubSpot's own model improvement. There is no model selector, per agent model choice, bring your own key option or routing policy, so control covers data access, not 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

An administrator sets the org-level model for Agentforce on the Agentforce Agents page in Setup, and Agent Script overrides it for a specific agent or subagent with a model configuration naming any supported model. The supported list spans Salesforce-managed models from OpenAI, Anthropic on Amazon Bedrock, Google Gemini on Vertex AI and Amazon Nova, most running inside the Salesforce trust boundary, and customers can also bring their own model.

Sourcedeveloper.salesforce.com/docs/ai/agentforce/guide/ascript-model.mdread 2026-09-18

Developers get a full platform at developers.hubspot.com, with REST APIs generated from OpenAPI specifications and the spec served at a public endpoint, maintained client libraries including Node, public apps that authenticate 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 with read and write access to CRM objects over OAuth with PKCE, managed through MCP auth apps, 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.

Sourcedevelopers.hubspot.com/ai-tools, developers.hubspot.com/docs/api/client-libraries and developers.hubspot.com/spotlightread 2026-10-01

The Agent API opens a conversation with an agent, sends prompts synchronously or as a server-sent event stream with progress, text-chunk and end-of-turn events, and closes the session, authenticated through a connected app with OAuth scopes, with a published Postman collection and Node.js client. Inside the platform, an invocable action runs any active agent from Apex or Flow, and Salesforce Hosted MCP Servers expose Salesforce data and automation to external AI agents and LLM applications over MCP.

Sourcedeveloper.salesforce.com/blogs/2025/04/build-headless-agents-with-the-agent-api.htmlread 2026-09-18

After a configuration is generated, the customer adjusts instructions, adds actions and tests the agent before publishing it. A/B testing on marketing email, custom reporting, workflow health monitoring on the higher tiers and the per-agent performance view in Agent Hub sit alongside. No score, pass rate, judge verdict or other comparable result comes out of that test, and no sandbox, replay, regression suite or evaluation harness over agent behavior is described.

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

Agents are tested before release with scored test cases. A test spec describes test cases as utterances with the expected agent behavior, can set context variables and include out-of-the-box metrics in the output, and is created and run in a development org from the command line or the Agentforce Testing Center, with pass and fail results.

Sourcedeveloper.salesforce.com/docs/ai/agentforce/guide/agent-dx-test.mdread 2026-09-18

Every action runs through a programmatic route. Agents act on the Smart CRM and reach third-party systems through integrations, the marketplace and APIs, and the developer MCP server drives an AI code editor, not 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

Salesforce does not describe an Agentforce agent operating a browser or desktop interface. Agents act through actions, flows, Apex, APIs, prompt templates and MCP tools, reached through the agent router and subagents in Agent Script.

Sourcedeveloper.salesforce.com/docs/llms-genai.txtread 2026-09-18