HubSpot MCP Server
@peakmojo
A Model Context Protocol (MCP) server that enables AI assistants to interact with HubSpot CRM data, providing built-in vector storage and caching mechanisms help overcome HubSpot API limitations while improving response times.
Overview
What is HubSpot MCP Server?
A Model Context Protocol (MCP) server that connects AI assistants directly to HubSpot CRM data, enabling interaction with contacts, companies, and engagement records. Built-in FAISS vector storage and caching improve response times and overcome HubSpot API limitations.
How to use HubSpot MCP Server?
Install via Smithery (npx -y @smithery/cli@latest install mcp-hubspot --client claude) or run the Docker image with a HubSpot access token. For Claude desktop, configure the server in mcpServers using the provided Docker command. The token must have scopes for contacts, companies, and sales-email-read.
Key features of HubSpot MCP Server
- Create contacts and companies with duplicate prevention
- Retrieve recent company activity and conversations
- Semantic search across retrieved HubSpot data
- FAISS vector storage for efficient retrieval
- Thread-level indexing with embedding caching
- Persistent storage between sessions
Use cases of HubSpot MCP Server
- Create HubSpot contacts and companies from a LinkedIn profile
- Monitor recent pipeline activity and conversation threads
- Automate CRM data entry without manual steps
- Query and summarize HubSpot engagement data via AI
- Enable AI assistants to manage HubSpot workflows
FAQ from HubSpot MCP Server
What HubSpot scopes does the access token need?
The token requires crm.objects.contacts (read/write), crm.objects.companies (read/write), and sales-email-read.
How do I install HubSpot MCP Server?
You can install via Smithery using npx, or pull the Docker image directly with docker run -e HUBSPOT_ACCESS_TOKEN=your_token buryhuang/mcp-hubspot:latest.
Where does persistent storage data live?
Data persists between sessions in a configurable storage directory. When running Docker, mount a host volume using -v /path/to/storage:/storage.
How is data retrieved from HubSpot?
The server provides dedicated tools (e.g., hubspot_get_active_companies, hubspot_get_recent_conversations) and a semantic search tool (hubspot_search_data) that uses FAISS vector storage.
What transport or authentication does the server use?
Authentication uses a HubSpot access token passed via the HUBSPOT_ACCESS_TOKEN environment variable. The server communicates over stdio when run via Docker CLI.
