HubSpot MCP server

Why Pathbound

Your customer doesn’t live only in HubSpot.

A direct HubSpot MCP shows your agent HubSpot. Pathbound shows your agent the customer — across HubSpot, Gmail, Intercom, your tracker, Apollo, Calendly, and the rest of your stack — through one unified profile served over the Model Context Protocol.

Beyond HubSpot

Gmail threads, Intercom conversations, website page views, Apollo enrichment, Calendly meetings — all stitched onto one timeline per contact. Your agent stops asking "who is this?" on every other tool.

Identity-resolved

Same person as [email protected] in HubSpot, [email protected] in Gmail, "Jane R." in Intercom — Pathbound resolves to one entity. A direct connector only knows its own IDs.

Built for LLMs

Token-efficient response shapes (count_only, fields=summary), server-side aggregation, batch tools. Agents get the answer, not raw JSON dumps. Stays inside the context window.

Audited writes

Update a contact once — Pathbound writes back to HubSpot, fans out to your other sources, and logs every change with actor, timestamp, and result. Reversible. The audit trail SOC 2 and GDPR auditors look for.

What it looks like

What does HubSpot data look like in Claude?

Connect HubSpot once. Your agent gets identity-resolved profiles, the full activity timeline, and write access — through one MCP server, identity-resolved with Gmail, Intercom, your tracker, and your other sources.

MCP

HubSpot, exposed as MCP tools.

Claude · pathbound MCP connected

you

Brief me on [email protected] — what's her HubSpot deal status, recent activity, and any risk signals?

↳ get_contact { email: "[email protected]" } 200 · 312 tokens · 480ms

contact Jane Reyes · Senior PM

company acme.com · fintech · 240

hubspot lifecyclestage = SQL · evaluation

↳ get_contact_timeline { contact_id: "ctc_01H..." } 8 events · 480ms

tracker visited /pricing · 2× today

gmail replied to follow-up · yesterday

hubspot deal_stage → presentation

claude

Jane is Senior PM at Acme (fintech, 240). HubSpot has her at SQL · evaluation; deal moved to presentation last week. She visited /pricing twice today and replied to your follow-up yesterday.

signal Active engagement, deal advancing. Lead with billing, not features.

claude_desktop_config.json

{
  "mcpServers": {
    "pathbound": {
      "url": "https://mcp.pathbound.ai/mcp"
    }
  }
}

HubSpot tools out of the box · read & write

get_contact — unified profile across HubSpot + your other sources

get_contact_timeline — page views, email replies, deal stage changes

list_contacts — filter by lifecycle stage, lead status, industry

aggregate_data — top contacts by activity, breakdowns by stage

create_contact_note — log call recaps, research, internal context

update_contact — patch lifecycle stage, lead status, properties — back to HubSpot

…plus 20 more — 26 MCP tools total, across contacts, companies, events, and conversations.

Setup

How do I connect HubSpot to Claude?

OAuth one connector. Drop a snippet into your MCP client. Ask a question.

01

Sign up for Pathbound

Create an account at pathbound.ai. Free during public alpha.
02

Connect HubSpot

In the dashboard, go to Integrations → HubSpot → Connect. OAuth via HubSpot's official OAuth API in one click; sync runs every 60 minutes and pulls contacts, companies, lifecycle stage, lead status, and custom properties.
03
Add Pathbound to your MCP client

Drop the snippet below into your Claude Desktop config (or in Claude.ai web: Settings → Connected Apps → Add custom MCP → enter https://mcp.pathbound.ai/mcp). The same URL works for ChatGPT, Cursor, Cline, and any MCP client that supports OAuth.
claude_desktop_config.json
```
{
  "mcpServers": {
    "pathbound": {
      "url": "https://mcp.pathbound.ai/mcp"
    }
  }
}
```
04

Run a test prompt

Ask the agent: “How many contacts do I have in HubSpot?” The agent calls list_contacts via MCP and returns the count.

FAQ

Common questions.

Is this an official HubSpot MCP server?

Pathbound is an independent platform. We connect to HubSpot via HubSpot's official OAuth API and expose your data through one Model Context Protocol server alongside your other sources. Not affiliated with HubSpot.

How is this different from the HubSpot REST API?

REST is for backends and scheduled jobs. MCP is for AI agents — it’s the protocol Claude, ChatGPT, and Cursor speak. Pathbound’s MCP also unifies HubSpot with Gmail, Intercom, your website tracker, and other sources into one profile per contact, so the agent doesn’t glue 6 APIs together itself.

Which MCP clients does this work with?

Claude.ai, Claude Desktop, Claude Code, ChatGPT (with custom connectors), Cursor, Cline, Continue — any client that supports the Model Context Protocol with OAuth.

Can the agent write to HubSpot, or just read?

Both. Pathbound exposes 16 read tools (list, get, timeline, conversations, aggregate, batch) and 10 write tools (create / update contacts and companies, create / update / delete notes). Writes sync changes back to HubSpot.

How does authentication work?

OAuth 2.0 end to end. The MCP client opens Pathbound’s login; you log in once and the agent gets scoped access tied to your user. Revoke anytime from Settings → Connected Apps.

Is my customer data secure?

Tenant-isolated and OAuth-scoped per user, with encryption at rest via our infrastructure provider's default configuration. GDPR-aligned; SOC 2 readiness in progress. Pathbound does not sell customer data, and we don't call AI providers on your behalf — your MCP client does, under your account and the AI provider's terms.

What about Pipedrive, Salesforce, Apollo, Intercom?

Same MCP server, same tools — just connect a different source. See the integrations directory.