Skip to main content
heyBTW’s MCP server is built for AI agent consumption. Once your workspace is connected, any MCP-compatible agent can query event intelligence, partner relationships, attribution, and pipeline outcomes in natural language. This page shows the patterns that work today, grouped by what an agent or its user is actually trying to do. Every query in the sections below maps to one of the eight live read tools; the ninth tool, activate_list, writes and is covered separately under The one write tool. Setup, authentication, and tool schemas live in MCP Tools.

​
Account and event intelligence

  • “What events has Chargebee attended in the last 6 months?”
  • “Show me engagement history for stripe.com.”
  • “Which accounts have attended 3 or more events and have open pipeline?”
These hit get_cross_event_patterns, with get_account_relationship for a single-account deep dive.
query_event_intelligence also answers these and is still served, but it is deprecated and being retired (HEY-1895). Prefer the two tools above in new work.

​
Partner and relationship intelligence

  • “Which partners have a relationship with OneTrust?”
  • “Who could give us a warm intro to Notion?”
  • “Which partners bring decision-makers versus practitioners?”
These hit get_partner_insights.

​
Attribution and pipeline impact

  • “Which deals can we attribute to the Berlin Fintech Dinner?”
  • “Show me Sourced versus Influenced pipeline from our Q1 events.”
  • “What’s the time-to-pipeline for event-sourced leads at MAU Vegas?”
These hit get_event_attribution. When reporting the results, represent attribution honestly: lead with Influenced, claim Sourced only with strong evidence, and always cite provenance.

​
Planning and invite lists

  • “Who should we invite to our Fintech dinner in Berlin next month?”
  • “Build an invite list for our SaaS roundtable.”
  • “Which accounts with open pipeline haven’t been to one of our events yet?”
These hit get_invite_recommendations.
This tool scopes to one event, named either by event_name (a human-readable name, fuzzy-resolved) or by channel_id. Beyond naming the event it takes no targeting parameters: there is no region, vertical, partner, budget, or ROI-target filter, and no account tier to filter on. Phrasing a request as though those exist will not narrow the result — name the event, then filter or segment the returned candidates in the agent.

​
Conference and multi-event reporting

There is no collection-level rollup tool on this surface today. An earlier version of this page pointed these questions at get_conference_summary, which was never a registered tool. Answer them by calling get_event_details once per event in the collection and aggregating in the agent, and say so rather than implying a single deduplicated rollup was returned.
  • “Summarize our performance at MAU Vegas across all activations.”
  • “Give me deduplicated attendance and ROI for our SaaStr presence.”
These fan out over get_event_details, one call per activation.

​
Combine heyBTW with your CRM

An agent can run heyBTW’s MCP alongside your CRM’s own MCP (Salesforce, HubSpot, or Attio) and join the two itself, without heyBTW ever touching your CRM data. This is how teams get pipeline-aware attribution while keeping CRM data in place. See Join heyBTW with your CRM through MCP.

​
How responses are structured

Every tool response includes a metadata envelope describing the data sources queried, the time range applied, the filters that were active, and a coverage note about data completeness. Agents can use this to cite their work and to flag when results are partial. When no data matches the query, tools return empty results with a clear coverage note. They do not error. Read the coverage note before treating a zero as real.

​
What the agent will not see

heyBTW is workspace-scoped. Every query returns only the data inside the workspace whose API key authenticated the request. Cross-company signals appear only when both sides have opted into a collaboration. Full data handling principles are in Data Handling.

​
The one write tool

Everything above is read-only. One tool on this surface is not: activate_list sends a saved List to its linked Luma event.
activate_list with action="confirm" sends real invites to real people, and it is not idempotent. Call action="preview" first, present the eligible count and the named target event to the human, and send only on their explicit go-ahead. Read when a retry is safe before re-issuing any confirm.
  • “Send the Berlin dinner list to the event.” → activate_list, preview first.
  • “How did the invites we sent last week land?” → activate_list with action="funnel".
Earlier versions of this page listed push_event_data, confirm_enrichment, and update_attendee_status here. None of the three is registered on this surface. See MCP Tools.