Operator-narrative review · Updated 2026-05-22

HubSpot MCP Review (2026): the broadest MCP surface in the CRM category — still Public Beta

HubSpot ships a hosted Model Context Protocol server covering the full platform — CRM, Marketing Hub, Sales Hub, Service Hub, Operations Hub, Content Hub, and the Breeze AI agent suite (Prospecting, Customer, Content, Social, Data). It's hosted by HubSpot, OAuth-authenticated, currently in Public Beta as of 2026. This is the broadest MCP surface in the CRM category — and the differentiator vs Close MCP and Attio MCP is breadth, not depth.

Quick context. We run StackSwap MCP — a GTM-focused MCP server. We're a HubSpot affiliate; the analysis below is the same one we'd give a friend evaluating HubSpot MCP against Close and Attio MCP cold.

Want to try HubSpot?

HubSpot ships the broadest native MCP surface in the CRM category

CRM + Marketing + Sales + Service + Operations + Content + Breeze AI agents — all on one MCP. Public Beta as of 2026; install the MCP and start using it now, deploy at GA for change-controlled environments.

Start with HubSpot →Affiliate link — StackSwap earns a commission if you sign up for HubSpot. We only partner with tools we'd recommend anyway.

What HubSpot MCP is — the breadth advantage

HubSpot runs a hosted MCP server at the platform level. You connect Claude Desktop, claude.ai, ChatGPT, Cursor, or any MCP-compatible client via OAuth. Once connected, the LLM sees the full HubSpot platform surface — not just CRM records, but Marketing Hub assets (forms, emails, landing pages, workflows), Sales Hub sequences and tasks, Service Hub tickets, Operations Hub data-sync and custom code, Content Hub assets, and the Breeze AI agent suite.

Structural distinction: Close MCP gives the LLM Close-shaped data. Attio MCP gives the LLM Attio-shaped data. HubSpot MCP gives the LLM the full GTM stack-on-one-platform — because HubSpot's product is the full GTM stack-on-one-platform. The breadth matters for cross-hub workflows: a single Claude conversation that reads Service Hub tickets, correlates them to Marketing Hub email engagement and Sales Hub deal stage, and drafts cross-team next-actions. That kind of synthesis is impossible without a unified MCP.

Separate from this hosted MCP is the local Developer MCP HubSpot ships for CLI and developer-environment work — that's a different server for a different use case (build custom apps, manage API auth, run local development workflows). Don't confuse the two; this review is about the platform-facing hosted MCP.

The Breeze AI integration is the second-order advantage

Breeze is HubSpot's bundled AI agent suite, repriced to outcome-based credits in April 2026: Prospecting Agent at $1 per qualified lead (100 credits), similar pricing across Customer Agent (AI service rep), Content Agent (marketing asset generation), Social Agent (post scheduling + drafts), and Data Agent (enrichment + cleanup).

The MCP layer exposes Breeze agent invocations as first-class tools. From a connected Claude session, you can have Breeze Prospecting Agent research and draft outreach to 50 accounts, then have Claude review the drafts and edit them before they queue for human send. The composition (Breeze runs inside HubSpot with full HubSpot context, Claude orchestrates across the broader stack) is the workflow shape that justifies running both an external LLM and Breeze in 2026.

The credit-burn warning applies here. An enthusiastic agent loop can burn through Breeze credits faster than expected. Mitigations: connect with a non-admin user with capped Breeze allocation, system-prompt a confirmation gate before any Breeze invocation, watch the Breeze credit dashboard for the first week. The visibility is sharper than Apollo's enrichment credit visibility — HubSpot exposes per-agent consumption in real time.

What you can actually do with HubSpot MCP

HubSpot MCP vs Close MCP vs Attio MCP

DimensionHubSpot MCPClose MCPAttio MCP
StatusPublic Beta (2026)GAGA (Feb 19, 2026)
Surface scopeCRM + Marketing + Sales + Service + Ops + Content + BreezeCRM + telephony + Chloe AICRM (relational object model)
AuthOAuth with HubSpot scopesOAuth 2.0 + Dynamic Client RegistrationOAuth only
Write gatingPer-operation OAuth scopesClose-Scope header (read / write_safe / write_destructive)Writes require confirmation in client
Bundled AIBreeze (Prospecting, Customer, Content, Social, Data) — outcome-pricedChloe AI (notetaker, drafts, enrichment, voice)None — bring your own LLM
Fits best whenMarketing + sales + service share contact graphInside-sales, phone-led, 20+ dials/dayRelationship-led, deep object model
Entry pricingFree CRM real; Starter $15-$20/seat; Pro $90-$890; Enterprise $150-$3,600+$9-$139/user/moIncluded in Attio subscription

The honest framing: pick the CRM by motion fit, then enjoy the MCP layer. HubSpot MCP's advantage is breadth (one MCP covers the full GTM platform); Close MCP's advantage is depth + Chloe; Attio MCP's advantage is relational object model + sharp write confirmation. All three ship serious native MCP — MCP capability isn't the differentiator. Motion fit is.

The Public Beta caveat — what to plan around

HubSpot MCP is in Public Beta as of 2026. Three honest implications:

The setup gotcha — scoped user, not admin

Same standard advice. Don't connect your HubSpot admin (Super Admin) account to Claude or ChatGPT. Create a separate HubSpot user with scoped permissions for AI work — read/write access only to the hubs and records the agent needs, capped Breeze allocation so credit-burn risk is bounded. The LLM inherits the user's permissions; an admin connection gives the agent full platform powers, which is overkill and risky during Public Beta.

Where StackSwap MCP fits

HubSpot MCP exposes HubSpot data. The cross-vendor question — "are we paying for hubs we don't use", "should we migrate Marketing Hub off HubSpot to Klaviyo or ActiveCampaign at our scale", "what's our actual GTM tooling spend" — sits at a different layer. That's where StackSwap MCP slots in: ~400 GTM tools, monthly costs, 104 hand-verified overlap pairs, swap recommendations. HubSpot MCP for "what does my HubSpot say", StackSwap MCP for "what should my GTM stack do." Both load into the same Claude session.

Want to try HubSpot?

HubSpot MCP is the broadest CRM-category MCP — install it during Public Beta and ship as it goes GA

Free CRM is real (unlimited users, no trial timer); paid hubs ladder from Starter $15-$20/seat/mo. The MCP layer covers everything you pay for.

Start with HubSpot →Affiliate link — StackSwap earns a commission if you sign up for HubSpot. We only partner with tools we'd recommend anyway.

FAQ

HubSpot MCP is HubSpot's hosted Model Context Protocol server covering the full HubSpot platform — CRM (contacts, companies, deals, tickets), Marketing Hub (forms, emails, landing pages, workflows), Sales Hub (sequences, tasks, calls), Service Hub (tickets, knowledge base), Operations Hub (data sync, custom code), Content Hub, and the Breeze AI agent surface (Prospecting, Customer, Content, Social, Data). It's hosted by HubSpot, OAuth-authenticated, and currently in Public Beta as of 2026. Endpoint docs at developers.hubspot.com/mcp. Separate from this is the local Developer MCP (also from HubSpot) for CLI and developer-environment work — that's a different server for a different use case.

Yes. The hosted HubSpot MCP exposes Breeze AI agent context across the full agent suite: Prospecting Agent (the autonomous SDR that researches and drafts outreach), Customer Agent (the AI service rep), Content Agent (generates marketing assets), Social Agent (schedules and drafts social posts), and Data Agent (enrichment + cleanup). The integration matters because Breeze is the AI surface inside HubSpot — connecting Claude or ChatGPT via MCP lets you orchestrate across both Breeze (inside HubSpot, with HubSpot data context) and your external LLM client (with the rest of your tooling). As of April 2026, Breeze pricing moved to credit-based ($1 per qualified lead for Prospecting Agent, similar outcome-based pricing across the agents).

OAuth — standard HubSpot OAuth flow with scopes. The LLM inherits exactly what the authenticated user can do in HubSpot: which CRM records they can read/write, which hubs they have access to, which Breeze agents they can invoke. The scope model is HubSpot's existing OAuth scopes (crm.objects.contacts.read, crm.objects.deals.write, marketing.workflows.read, etc.) — no MCP-specific permission system layered on top. Audit logging is standard HubSpot — every action shows under the authenticated user in the activity history. Compared to Close's Close-Scope header model (read / write_safe / write_destructive) which gates writes at the session level, HubSpot's scope model is per-operation and inherited from the standard HubSpot OAuth setup.

Realistic workflows: (1) ask Claude to summarize this week's marketing-to-sales handoff — which MQLs converted, where the funnel leaked; (2) read contact lifecycle stage + recent activity + deal context to draft personalized re-engagement emails for stalled deals; (3) orchestrate Breeze Prospecting Agent invocations from Claude — "have Breeze research and draft outreach to these 50 accounts, queue the drafts for review"; (4) bulk-update contact properties, deal stages, or workflow enrollments via natural-language prompt; (5) cross-hub queries that span Marketing + Sales + Service in one conversation ("which paying customers have open service tickets and stalled renewal opportunities"); (6) Content Hub asset generation — describe a landing page, Claude drafts the copy and creates the page in HubSpot under draft status. The Public Beta status means the surface is iterating; check the developers.hubspot.com/mcp changelog for the current capability set.

HubSpot MCP is structurally the broadest of the three — covers Marketing, Sales, Service, Operations, Content, and Breeze AI agents on one MCP. Close MCP covers the CRM + telephony + Chloe AI inside-sales-execution surface. Attio MCP covers the relational object model CRM. Pick by motion fit: HubSpot when marketing + sales + service share contact records and the motion depends on lifecycle nurture + cross-team attribution; Close when phone-led inside sales is the engine; Attio when relationship-led with deep object models. The MCP layer follows the CRM decision, not the other way around — all three ship serious native MCP, so MCP capability isn't the differentiator. HubSpot's MCP advantage is breadth; the tradeoff is the per-hub seat-padding that hits as you move past Sales Hub Starter and into Marketing Hub Pro.

Breeze agents run on credit-based pricing as of April 2026: Prospecting Agent at $1 per qualified lead (100 credits), similar outcome-based pricing across the suite. When Claude invokes Breeze Prospecting Agent via MCP, those credits consume from your HubSpot Breeze allocation — and an enthusiastic agent loop can burn through credits faster than expected if you don't set guardrails. Mitigations same as Apollo MCP: (1) connect with a non-admin HubSpot user with capped Breeze allocation; (2) system-prompt a confirmation gate before any Breeze invocation; (3) watch the Breeze credit dashboard for the first week of MCP use. The visibility into agent-credit consumption is sharper than Apollo's enrichment-credit visibility, which helps once you're calibrated.

Yes, with Public Beta caveats. The shipped security model is OAuth-based with user-scoped permissions (standard HubSpot OAuth scopes), full audit logging under the authenticated user, and HubSpot infrastructure (which is enterprise-grade — SOC 2, GDPR, the usual compliance). The Public Beta status means: (1) the operation catalog is iterating, so workflows may shift; (2) write-confirmation UX depends on your MCP client; (3) Breeze credit consumption needs operator-side guardrails. For SMB and mid-market production deployments, the shipped MCP is safe. For enterprise compliance with strict change-control rules, treat the Public Beta status as 'evaluate now, deploy after GA' and pin which capability set you're depending on.

For 2026 CRM evaluations, MCP support is moving from nice-to-have to structural advantage for AI-curious operators. HubSpot's broad-platform MCP is a real differentiator vs CRMs that don't ship MCP — but the bigger HubSpot eval questions (per-hub pricing math, Marketing Contact tier escalation, the cliff between Pro and Enterprise) don't change because of MCP. The framing: if you're already on HubSpot and AI-curious, install the MCP and start using it. If you're shopping for a CRM in 2026 and HubSpot fits your motion (marketing + sales + service sharing contact records), native MCP is now a structural plus. If your motion doesn't fit HubSpot (phone-led inside sales, relationship-led with deep object models, enterprise governance), the MCP layer doesn't make HubSpot the right pick — pick by motion fit first.

Related reading

Canonical URL: https://stackswap.ai/hubspot-mcp-review. Disclosure: StackSwap is a HubSpot affiliate.