Chat with your HubSpot data

This n8n template gives you a chat-style assistant that can search your HubSpot CRM on demand. Ask natural-language questions like “show me leads in Germany” or “what deals close next month,” and the agent translates your request into precise HubSpot searches—then answers in plain English.

Great for founders, AEs, and ops folks who want quick answers without clicking through the CRM.

How it works

• Chat trigger starts a session from your n8n chat UI or embed.

• AI Agent (Gemini 2.5 Pro) interprets the message and:

  • Chooses the right HubSpot search (contacts or deals).
  • Fills filter property, operator (EQ, NEQ, GT, GTE, LT, LTE, BETWEEN, IN, NOT_IN), and value(s).
  • Requests specific properties (email, name, lifecycle stage, owner, activity timestamps, etc.).

• HubSpot tools execute live queries:

  • Contacts: flexible property filter + free-text query.
  • Deals: filters by owner and core deal fields (stage, amount, pipeline, close date).

• Memory buffer keeps the last turns so you can say “now only show closed won over 10k” and the agent understands context.

How to use

1. Connect credentials

   • HubSpot OAuth on both HubSpot Tool nodes.
   • Google Gemini API key on the Gemini Chat Model.

2. Open the chat (the “When chat message received” node).

3. Ask questions like:

   • “Find contacts named Hans created after Sept 1.”
   • “Deals owned by me in Proposal with amount > 10,000.”
   • “Contacts with lead status = New and no email reply in the last 14 days.”

4. Refine with follow-ups:

   • “Sort by most recently contacted.”
   • “Only Germany.”
   • “Show top 5 with emails.”

Requirements

• HubSpot (OAuth2)
• Google Gemini (API key)

Notes & customization

• Property/operator control: The contact search node lets the agent set both the property (e.g., email, lifecyclestage, hs_lead_status) and the operator (EQ, IN, BETWEEN in epoch ms for dates, etc.).
• Owner filtering for deals: Uses hs_all_owner_ids; swap or extend to filter by pipeline/stage ranges.
• Guardrails: Add allowlists for searchable properties or cap result counts to avoid noisy answers.
• Display format: Have the agent return concise tables (name, email, stage, last activity, CTA).
• Handoffs: Add Slack/Email actions—e.g., “post this list to #sales” or “export to CSV.”
• Telemetry: Log queries for later dashboards (common searches, coverage gaps).

Troubleshooting

• No results? Loosen operators (use IN lists, broaden dates) or include a free-text query.
• Date filters: Provide epoch ms for GT/GTE/LT/LTE/BETWEEN on time fields (the agent handles this; keep system time in UTC).
• Too chatty? Reduce memory window or ask the agent to summarize to bullet points.