Live demo — UserActivity.ai

A problem, found and fixed — then confirmed.

This is the part that doesn’t exist anywhere else: an AI agent that watches real user behavior, fixes what’s broken, and checks its own work — all in one conversation.

Fig. A — The loop in 90 seconds. The step-by-step session is below.

The agent connects

One MCP entry in your project config. Nothing to install — your coding agent talks straight to the hosted API.

// .mcp.json — Claude Code, Cursor, Windsurf…
{
  "mcpServers": {
    "useractivity": {
      "type": "http",
      "url": "https://useractivity.ai/api/mcp",
      "headers": { "Authorization": "Bearer sk_live_…" }
    }
  }
}

✓ Connected to useractivity — 14 tools available

You ask. It diagnoses.

Plain language in. The agent queries the behavioral data and reads the problem back to you — no dashboards to learn, no event taxonomy to design.

You:    Signups dipped this week — what's going on?

Agent:  [ calls get_frustration_signals({ page: "/signup" }) ]

        /signup is scoring 38/100 on frustration:
          • 1 in 5 sessions rage-click the submit button
          • the button throws a validation error intermittently
          • 64% never get past the first field

        Looks like a race in the form validation. Want me to fix it?

It fixes — and records the change

The agent edits the code, then marks the change so the impact can be measured later. That marker is what makes the next step possible.

Agent:  [ edits SignupForm.tsx — await validation before submit ]

        [ calls record_change({
            note: "Fix signup validation race",
            ref:  "a1b2c3d"
          }) ]

        ✓ Change recorded. I'll check the impact once data comes in.

It confirms the fix actually worked

the loop closes

A day later, the agent measures before vs. after the change. This is the part no other analytics tool does — closing the loop between what you shipped and what users did next.

You:    Did the signup fix work?

Agent:  [ calls get_change_impact({ pivot: "a1b2c3d" }) ]

        Yes — comparing the 2 days before and after the change:
          • Frustration on /signup:  38 → 71   (+33)
          • Rage clicks:             −82%
          • Signup completion:       +23%

        The fix worked. The loop is closed.

That fourth step — confirming the fix worked — is the loop. Collecting behavior is table stakes; closing the gap between a change and its effect, inside the agent conversation where you already work, is the thing built for how you actually ship.

And there’s a dashboard, too.

The agent surface is the point — but a human can still see everything at a glance. This is the real Overview, populated with sample data from a fictional store. Not a real customer’s account, and never an empty shell.

useractivity.ai/dashboard

Sample data

Northwind Store

Betashop.northwind.example7d

4,182

Visitors

11,540

Pageviews

Last event 4 minutes ago

Mon

Tue

Wed

Thu

Fri

Sat

Sun

0006121823

Health

Composite · 7d

Reading healthy overall, with two areas to watch this week.

Frustration64

Scroll75

Sessions78

Navigation70

Engagement66

Forms81

Top issues

·Engagement (66/100) — Median time on /products fell to 1m 48s, down from 2m 30s last week — visitors are bouncing before they reach the grid.
·Frustration (64/100) — Dead clicks on the /cart promo banner — 9% of sessions tap it expecting a link, but it isn't interactive.

Elevated signals

Engagement

/products engagement is sliding — worth a look this week

Navigation

A new /products → /cart → /products u-turn is forming

Forms

Checkout completion climbed to 81%, up 6 points

Top pages

Path	Views	Visitors
/	4,310	2,980
/products	3,120	1,840
/signup	1,640	1,290
/pricing	1,180	910
/cart	1,290	760

Recent changes

Fix signup validation race
2 days ago
Agenta1b2c3d✓ Frustration on /signup 38 → 71 · completions +23%
Deploy v2.4.0 — checkout refactor
2 days ago
Deploy9f4e1c0
Launch: spring product drop email
5 days ago
Campaign

Fig. B — The real dashboard, on sample data. Dark because that’s how the product ships.

See the loop close.