Analytics · email · in-app guides · one agent running them
All the growth work that keeps slipping. Shipped.
Loop finds why your numbers dropped, drafts the fix, ships it to exactly the right users and proves whether it worked. You approve each step. No analyst, no sprint, no new hire.
Free plan, unlimited seats. Engineer time required: about two minutes, once.
The loop, running: one circuit, one gate
Sound like Monday?
“Activation dropped and I don't know why. The data team can look at it in two weeks.”
LoopAsk in plain English. Answer in a minute, not a sprint.
“I have twelve growth ideas and zero engineering time for any of them.”
LoopShip the fix yourself. No ticket, no sprint, no engineer.
“We shipped the onboarding change. Did it work? Honestly, nobody knows.”
LoopEvery change ships with a holdout. You always get a verdict.
The agent's room; the lights stay down while it works
Describe the outcome. Approve the fix.
“Activation dropped this week. Find out why and fix it.”
That's the whole brief. This is interactive: the one blue step is yours, and it cannot be skipped.
The agent · diagnosing
“Activation fell 9 points this week. People are getting stuck at ‘connect your data’, and trials from last week's campaign hit it hardest.”
Signed up3,207
Connected data−63% leave here1,186
Activated912
the drop from the top of this page; the agent is on it
41,209 sessions, 14 days. It reads the same funnels you can open yourself, and every claim links to the query it ran. When confidence is low it hedges in words, never with decoration.
Ship and Measure are locked until you approve. Not for effect; the agent can't skip the gate either.
Three tools. Three bills. Still no answers.
Your analytics tool knows who got stuck. Your email tool can't hear it. Your onboarding tool can't prove anything. Between every insight and every action sits an export, a CSV and a week of your life.
- Product analyticssubscription #1
↓ export.csv, by hand, again
- Lifecycle emailsubscription #2
↓ audience rebuilt, by hand, again
- In-app guidancesubscription #3
↓ measured by nobody
Three subscriptions. Three seat bills. Three integration projects. Three copies of “activated” that never agree.
Loop
One bill. One audience that's true everywhere. One agent running all three, and measuring every move it makes.
- analytics
- guides
- experiments
$0 to start · $199 flat at scale. We won't print competitors' prices without a source. Check your own invoices; that's the comparison that matters.
Switching? Bring your history. Loop imports your event archive, so day one starts with your last two years, not a blank chart.
For founders, the same sentence in money: no analyst hire, no growth engineer, one subscription instead of three. Live today, not next quarter, and your growth backlog stops being a backlog.
Stalled trials
defined once · 3 editable rules1,204 people · live
The audience the agent built in the run above: the same object, not a copy. Everything below reads it.
- Emailday-2 nudge → reads this audience
- Guidein-app checklist → reads this audience
- Experiment50/50 → scored on the same people
Audiences
Stop rebuilding the same audience in three tools every week.
Define “stalled trials” once. The email reads it. The in-app guide reads it. The experiment reads it. Change it once and it changes everywhere, and the same numbers score the result, so the three tools can never disagree again.
Answers
Ask in plain English. Get the answer over coffee.
Type the question the way you'd ask a colleague. Loop shows the chart, words the answer, and shows the exact query it ran. And because it captures everything from day one, you can ask questions you didn't know to track; the data is already there.
No tracking plan. No schema meeting. No SQL.
“The drop is at ‘connect your data’, and it lost you 63% of last week's trials.”
Visited pricing8,412
Signed up3,207
Connected data1,186
Activated912
Drill into the 1,186 who stopped here; they become an audience in one click.
Guides & tours
Tours that exist for a reason, and that prove they worked.
Checklists, tooltips, modals, banners, inline embeds, and multi-step tours that walk a user through a real flow, anchored to your actual UI, inside your product.
The agent can author one from a real drop-off and ship it to the exact segment that's stuck, after you approve it. Nobody else's tour tool knows why it exists.
- Ships as a 50/50 experiment, like everything else, so you learn whether the tour worked
- Dismissal is a right: always visible, comfortably tappable, permanently respected
- Wears your brand, computed from one brand colour and provably AA, never ours
Data sources
Connect data sourceConnect your data source
Point Acme at your warehouse and your first report builds itself. Takes about a minute.
the teal is Acme’s; Loop’s clay never appears inside your product
“Inconclusive is a dignified state, and it gets no confetti.”
Verdicts
Never again ship a fix and wonder if it worked.
Everything ships with a holdout, so everything gets a verdict: a lift with its confidence interval, or “inconclusive”, said out loud.
You've been sold “AI-powered” before, and it was a chat box that made things up. Loop shows its evidence, states its uncertainty, and is built so it can't dress a maybe as a win.
+4.2%
checkout conversion, not yet significant, and Loop says so
95% CI −0.8 … +9.2 · n 5,113 · still measuring, not dressed as a win
When a result is significant, like the checklist in the run above, it ships wider with the same treatment. When it isn't, it says so and keeps measuring. There is no version of this where a maybe becomes a win.
significance is derived from the interval, never from enthusiasm
Why you can trust it near your customers
The agent is powerful. That's exactly why it's leashed.
It can email your customers and change your product, so the leash is enforced in code, not in a prompt it could ignore. In writing:
The agent can
- Read your funnels, audiences and results, and cite them
- Draft emails, guides and experiments; drafts touch nobody
- Ship what you approved, to exactly who you approved
- Pause itself, at your spend cap or when a run looks wrong
The agent cannot
- Send anything to a customer without a named human pressing approve
- Spend past the daily cap you set; it stops and tells you
- Delete data; deletion is human-only, typed, and states what dies
- Call a maybe a win; honesty is built into the components themselves
One switch turns it off
Available to every role, always. Loop without the agent is still all three tools.
Every action leaves a receipt
What ran, what moved, who approved, with revert sitting right beside it.
Consent-aware from the start
Loop holds events until your users say yes, and deletion means deleted.
Locked to your site
A key copied from your page is useless anywhere else. Your data can live in your cloud, on your domain.
Setup
Your engineer needs two minutes. Once.
One line on your site, and Loop starts learning immediately: every page, click and signup, captured without a plan. You'll watch your first event land live, and be looking at a real funnel over coffee.
No tracking plan. No schema meeting. No SQL. This is the only code on this page, and the last time you'll be blocked on a sprint.
Paste one line; your key is already in it
It goes in your <head>. That's the whole engineering task.
<script src="https://cdn.loop.app/loop.js" data-key="wk_live_…" async></script>Watch your first event arrive, live
The console shows it land, honestly: no fake dashboard appears before real data does.
That's the whole install
There is no step three. The agent takes it from here, and you take the approvals.
Engineers: frameworks, keys and self-hosting are all in the docs.
One bill. Three tools retired.
Unlimited seats on every plan. You pay for data, not for people.
Explore
$0
no card, no clock
- 50,000 events a month
- Every surface, and the whole agent
- Unlimited seats
- Agent spend cap starts at $5/day
Growth
$199/month, flat
less than any one of the three tools it replaces is likely costing you. Check your invoices
- 10M events a month, then $20 per extra 1M
- Unlimited projects and seats
- Agent actions at cost; the cap is yours
- History import · roles · audit log
Self-hosted
Custom
Loop in your cloud, on your domain
- Your data never leaves your infrastructure
- First-party, unaffected by ad-blockers
- Retention without limits
- Migration help, shared channel
Honest answers
Can I turn the agent off?
Yes: one switch, available to every role, always. Loop without the agent is still full analytics, messaging and guidance. The agent is a colleague, not a hostage-taker.
Do I need a credit card to start?
No. The free plan has no card and no clock. It is not a trial that expires, it is a plan you can stay on.
What happens if I hit the event limit?
We never drop your data silently. On Growth, overage bills at $20 per extra 1M events. On the free plan we warn you at 80% and pause new ingestion at the limit; your existing data stays queryable.
Do you train models on my data?
No. Model calls run inference only. On a self-hosted deployment every call stays inside your own infrastructure.
What does leaving look like?
Export everything (events, users, segment definitions) in open formats, any time. Deletion is a typed confirmation that states exactly what is destroyed, and it means deleted.
Run your whole growth loop. Yourself. This week.
Loop is new, and that's why the whole product is on this page. Try it on your own data and judge us there.