Insights & MCP
Once data is flowing, the question shifts from "is this set up?" to "what is it telling me?", where's the dropoff, which channel converts, who's churning, what to fix this week. Two paths.
- This page — hand the funnel to your agent for ad-hoc, conversational questions over MCP.
- Reports — generate a saved, shareable AI analysis from the dashboard.
Use both. They share the same underlying model and data.
Hand it to your agent
Your agent connects to the FunnelFizz MCP server and calls read tools that return live funnel data. It then synthesizes, bottlenecks, comparisons, recommendations, without you opening the dashboard.
Paste any of these into Claude / Cursor / ChatGPT (assuming MCP is connected, see Getting Started → Path A):
Run a weekly review across all my funnels. Surface the biggest
dropoffs, channel that's underperforming, and one thing to fix.
Compare conversion across my three funnels for the last 30 days.
Which performs worst, why, and what would you change?
Pull the journey for the profile that just signed up. Where did
they come from, what did they read, what convinced them?
Why is my trial-to-paid conversion lower this month than last?
What's the dropoff between AWARENESS and CONSIDERATION on my main funnel?
The agent picks the right MCP tool for each question and stitches results into a narrative.
What the agent has access to
| Capability | Plan | Tool |
|---|---|---|
| Funnel overview (stage counts, conversion rates) | HOBBY+ | get_funnel_overview |
| Stage metrics with optional split-by | HOBBY+ | get_funnel_metrics |
| Side-by-side comparison across funnels | HOBBY+ | compare_funnels |
| Cross-funnel rollup + top funnels | HOBBY+ | get_workspace_overview |
| Search profiles by email/name | PRO | search_profiles |
| Latest 100 events for a profile (PII redactable) | PRO | get_profile_journey |
| Generate or fetch a saved AI report | HOBBY+ | generate_funnel_report, get_funnel_report |
Pre-baked review prompts
Built-in workflows your agent can run by name:
| Prompt | What it does |
|---|---|
weekly_review | Multi-funnel summary, last 7 days vs prior 7. |
diagnose_funnel | Pinpoints why one funnel is underperforming. |
find_dropoff | Locates the biggest conversion leak. |
Just say "run weekly_review", the agent invokes the prompt and surfaces results.
What the agent can't do
- Send email or trigger automations.
- Edit funnels, stages, splits, or conversions.
- Mutate or delete profile data.
- Anything outside the read scope.
So a "what should I change?" answer is always a recommendation, you make the change.
→ Setup: Hand it to your agent · MCP reference: /ai-agents/mcp
When to use Reports instead
Reports are the same kind of analysis, captured as a saved artifact. Use them when:
- You want a weekly funnel diagnostic you can revisit and forward to a co-founder.
- You're not in an agent session and just want to click "Generate report" in the dashboard.
- You want to export the analysis as Markdown for paste into Notion / a Loom doc / another model.
For one-off questions ("what's my worst stage?") MCP is faster. For artifacts you want to keep, generate a Report.
Related: Reports · Stages · Conversion · MCP