The marketing command center
Where the traffic comes from, which source produces leads that actually close, what a lead costs, whether the phone was answered — and the revenue at the end. One event stream from ads to CRM to calls to Stripe, with an advisor that builds the first dashboards from your own data.
In short
- Every stage reports into one pipeline: ad spend (Meta, TikTok), traffic (GA4), search (GSC), leads (Meta Lead Ads, GoHighLevel), calls (Twilio), revenue (Stripe).
- One unit everywhere — money in cents, events in one envelope — so CPL, CPA and ROAS are divisions, not export projects.
- The advisor profiles your data and proposes strategic dashboards with reasoning and evidence; one click creates, everything stays editable.
- Deterministic anomaly alerts watch every stage hourly — a missed-call spike or a lead drop emails you with the numbers.
- Your own LLM can run all of it by prompt over MCP, with a scoped key: 35 tools, from connect-a-source to read-me-the-revenue.
The funnel, stage by stage
| Stage | Source | Events | The question it answers |
|---|---|---|---|
| Spend | meta-ads, tiktok-ads | ad.spend_reported | what did attention cost today |
| Traffic | google-analytics, google-search-console | web.traffic_reported, seo.performance_reported | who arrived, from where |
| Leads | meta-leads, gohighlevel | lead.created, lead.booked | which channel produces leads that book |
| Calls | twilio | call.received, call.completed, call.missed | did the phone get answered when the ads ran |
| Close | stripe, gohighlevel | deal.created, payment.succeeded | what actually became money |
Ways in, for whatever your stack is
Set up by conversation — your LLM, our MCP
The whole surface is promptable. Connect any MCP-capable assistant to
@hookget/mcp with a scoped API key, and the setup conversation looks like this:
you: "Look at my HookGet data and tell me what my funnel is missing."
llm: → describe_data
"Spend, leads and revenue are flowing; calls and traffic are not.
Your funnel is blind between the ad click and the CRM entry."
you: "Connect my GA4 property 123456789 and build what you recommend."
llm: → create_source {provider: "google-analytics", config: {property_id: "123456789"}, ...}
→ suggest_dashboards
"The advisor proposes a Marketing command center (spend + leads +
revenue are all present). Reason: per-platform consoles each see one
stage; this view sees the hand-offs. Create it?"
you: "Yes."
llm: → create_dashboard {definition: …} # validated whole; scoped to YOUR project by the keyThe security shape, stated: the model proposes; HookGet's validated, tenant-scoped layer decides. A dashboard definition cannot name a tenant, queries compile server-side, and the MCP key carries exactly the scopes you gave it. Your prompts run on your AI account — HookGet adds no LLM, no per-call cost, no data sharing.
Questions
What does "the advisor" actually do?
It reads what your project's last 30 days actually contain — which event families are flowing, at what volumes, with which metrics — and proposes the dashboards that data supports, each with its reasoning and the evidence that fired it. When spend, leads and revenue all flow, it composes a command center for your exact mix (the call widgets appear only if telephony reports). Every proposal is validated before you see it; creating one is a click. It is deterministic: the same data always produces the same advice.
Where does AI fit, honestly?
Your AI, not ours. Connect your own LLM — Claude, or anything MCP-capable — to HookGet's MCP server with a scoped API key, and it can do everything this page describes by conversation: inspect your data, narrate the advisor's proposals, adapt them, create the dashboards, wire new sources, read the numbers back. HookGet's side stays deterministic (ADR: the model proposes, the validated layer decides), which means no LLM runs on our servers, nothing is added to your bill, and your prompts never leave your own AI account.
Which CRM and payment systems fit?
Out of the box: GoHighLevel (leads, deals, appointments), Meta Lead Ads (Facebook and Instagram forms), Stripe (payments through to disputes), Twilio (calls and SMS). HubSpot and generic CRMs connect today through the webhook door or an n8n flow; native connectors follow the same template. Any payment provider that signs webhooks can enter through the verified generic door.
What can I NOT see here?
Click-level attribution across networks — each ad platform reports its own attributed numbers, and HookGet lands them side by side in one unit rather than re-attributing. Session recordings and page heatmaps are a different product category. And LinkedIn Lead Gen needs their partner programme — today that route is LinkedIn → n8n/Zapier → the n8n source.