AM-135 · for Kite's trial-to-paid work (AM-106 credit ladder, AM-127 sign-ups)
How agentic products get free users to pay
We signed up for 11 agentic products and screenshotted every step from landing page to checkout (no card ever entered). Throwaway inboxes got four all the way through (Native, Okara, Sintra, Zapier Agents). Nikhil then signed a real Google Workspace account into a Composio cloud browser, which got five more through: Lindy, Tasklet, Gumloop, Genspark and Manus. Viktor (needs a Slack workspace) and Town (sign-up is a full inbox grant) are still documented from pricing pages and public docs. The products that converted us fastest did the work first: a URL or a work-email domain alone produced on-brand posts, a marketing brief or tailored suggestions within minutes, and the paywall then sold more of that work.
ComparisonPatternsStandout tacticsRecommendations for KiteBlocked sign-ups
Comparison
Click a product for its screenshot-by-screenshot flow. Prices as shown on 28 Oct 2026 in USD (some checkouts converted to INR from the India IP).
| Product | Pricing | Trial / free entry | Card? | What free allows | Paywall trigger | Key upgrade tactic |
|---|---|---|---|---|---|---|
| Native named · 37 steps Full flow |
Pro $79/mo ($47/mo first year on annual), Max $249, Agency $399 | No trial. Free = 3 posts to try from ~50 pre-generated for your brand | No to start; yes for anything past 3 posts | Approve 3 posts; every create/memory action is paywalled | Approving the 3rd post → '47+ more posts, made for Kite. Pick a plan to unlock them' | Sells your own already-made posts, shown blurred, with yearly preselected and a Halloween countdown |
| Lindy named · 21 steps Full flow |
Team $29.99/seat/mo for 3k credits (slider to $999.95); in-app Plus/Pro/Max | 7-day trial, 10,000 credits, no card; then a Free Monthly plan | No | Trial credits for real work (one research task = 10.7 credits); model choice locked to 'Auto' | 'Upgrade to choose a model' lock; trial end; [docs] credits empty → work pauses | Home suggestions personalized from the email domain alone; cost shown under every reply; $50 / 30% referral card |
| Viktor named · 17 steps Blocked |
Team $100/mo for 40k credits, flat $25 per 10k up to $95k; Enterprise | Up to $100 credits, never expire, every feature unlocked | No | ≈ 80–400 tasks; only top-ups, rollover and full rate limits are locked | Credits gone → Viktor pauses the task in Slack: 'Upgrade plan' / 'Ask an admin' | Out-of-credit asks travel up the org: Viktor DMs admins that '2 teammates are waiting on me' |
| Town named · 17 steps Blocked |
Starter $15, Pro $49 (annual), Power $99, Power Plus $199; Team $59/seat | 7-day Pro reverse trial at sign-up (or until trial credits run out), then Free | No | Trial = full Pro; after it, 30 chats/month, weekly briefing, routines pause | Trial ends → Free and routines pause; 80/90/100% credit emails; 'Out of credits' → upgrade | Reverse trial: you live on Pro for a week, then lose the automations you set up |
| Okara AI CMO added · 20 steps Full flow |
AI CMO Lite $129/mo ($108 annual), Pro $249 ($208 annual) | Free plan (documents) + 7-day trial, card required, defaults to annual | Yes for the trial (₹0 today, then the annual plan) | Product brief, 30-day strategy, competitor list, SEO audit; all 8 agents locked | End of onboarding → full-screen 'Start your free trial now'; every lock icon routes there | Free strategy, paid execution, with an Agent Log that narrates which steps 'unlock on the paid plan' |
| Sintra added · 29 steps Full flow |
'Sale': $48.50 first month (renews $97) or $15.60/mo first year (renews $624/yr); in-app $37.50/mo for 250 credits | Pay first + 14-day money-back; a hidden free 20 credits/mo via the login page | Yes on every website CTA; no via the hidden path | 20 credits/mo (≈ 6 research tasks); a 15-s video costs 51 | 'Generate video · 51' → 'You're out of credits — top up'; 'Low · 3 / 20' pill | Price shown on the expensive button + credit tiers translated into outputs ('up to 250 social posts / 5 videos') |
| Manus added · 18 steps Full flow |
$20 / $40 / $200 per month (annual −17%); team calculator defaults to 20 seats | Free: 1,000 + 300/day credits on the Lite model; 50%-off first month offered during onboarding | No | 1,300 credits; one research task = 235 credits; premium models locked | Plan picker inside onboarding; 'You're using 2.0 Lite' popup; post-task 'free Manus 2.0 trial for this message' banner | One secret premium answer, then 'Upgrade to continue receiving the most refined… responses' |
| Genspark added · 15 steps Full flow |
Plus $19.99/mo annual ($24.99 monthly), Pro $199.99; Team $35/seat (10-seat default) | Free 100 credits/day until a lifetime cap | No | 100-credit 'Daily Gift' (one research task = 28 credits); lifetime cap per docs | A 'Free weekly usage is here' gift button that opens the paid plan picker; composer 'pay once, not 3x' tab | Gift-framed upsells with zero onboarding; dollar-savings badges on annual |
| Tasklet added · 19 steps Full flow |
Starter $25/mo (10k credits), Pro $100, custom in $5 steps; no seat fees | Free plan, ~1,100 starter credits, plus credits for each setup milestone | No | ~1,100 credits + up to 3,750 milestone credits; per-thread daily cap | Header 'Upgrade' from screen one; credits run out (not reached) | Credits as activation rewards: +250 connect an app, +500 automate a task, +1,000 text or share the agent |
| Gumloop added · 17 steps Full flow |
Pro $37/mo for 20k credits ('7,400 + 12,600 bonus'); Enterprise | No free tier: onboarding ends at a 14-day trial that needs a card | Yes | Nothing without the card-backed trial | Last onboarding screen: 'Start free 14-day trial' or 'Sign out' | Opt-out trial as the only door; Google sign-up also asks for Gmail send + Calendar edit up front |
| Zapier Agents added · 27 steps Full flow |
Agents Pro $50/mo ($33.33 annual) for 1,500 activities; separate from Zapier platform plans | Free forever: 400 activities/month; sign-up also auto-starts a 14-day platform Pro trial | No | 400 activities/mo and 10 activities per run; one research task ≈ 25 activities | Mid-run, the agent itself: 'reached the limit of 10 activities per run on Agents' Free plan. Upgrade… Shall I continue?' | A per-run depth cap pitched by the agent at the moment of value (free users can still continue) |
Patterns across products
Do the work first, then charge for more of it
The two fastest 'aha's needed only a URL. Native returned ~50 on-brand posts for Kite in about a minute; Okara wrote a product brief and a 30-day strategy during a 3-screen profile. Both ran the analysis while asking their questions, and the paywall arrived when the user already had something in hand. Sintra did the opposite (15 typed setup screens before the first chat) and felt like work.
Credits everywhere, but the good ones speak in outcomes
Every product meters in credits or tokens. The ones that are easy to reason about translate them: Sintra's plan picker ('up to 250 social posts / 140 images / 5 videos'), Native's token table ('10 tokens a post'), Lindy's task-size bands, Viktor's '40,000 credits ≈ 80–400 tasks', Manus's example tasks with their credit cost.
Four ways in, ranked from generous to strict
(1) A large no-card grant with everything unlocked: Viktor ($100), Lindy ($50 for 7 days), Town (7-day Pro reverse trial). (2) Drip-fed free credits: Manus (300/day), Genspark (100/day to a lifetime cap), Sintra (20/mo), Tasklet. (3) A tiny free taste plus a hard paywall: Native (3 posts), Okara (documents only). (4) A card up front: Okara and Gumloop trials auto-convert; Sintra sells first with a money-back guarantee.
Paywalls sit on the user's own work
The sharpest upgrade prompts show what the agent already made or would make for this company: Native's blurred '47+ more posts, made for Kite', Okara's locked-agent modal over a blurred table of example work, and Sintra's half-built video board priced at 51 credits. Generic feature lists come second.
Annual is the default, with intro prices
Native (−40% + countdown, 'first year, then $63/mo'), Okara ('2 months free'), Town and Genspark (−20%), Manus (−17%) and Sintra (−70% first year, renewing at $624) all preselect annual. Native and Sintra make year one cheap and the renewal dearer. Native states the renewal price right on the card; Sintra keeps it in the checkout fine print.
Usage warnings escalate, and the agent delivers them itself
Sintra: a small counter, then a 'Low · 3 / 20' pill, then an 'out of credits' toast. Town: emails at 80/90/100%. Gumloop: emails at 75/90%, then pay-as-you-go. Viktor: 70% and 90% DMs with an estimated run-out date, then pausing the task in Slack and asking the admin. In-thread products (Viktor, Lindy) send these notices from the agent, in the conversation. Zapier's agent stops mid-run at the free plan's 10-activities-per-run cap and pitches the upgrade itself.
Habit hooks via daily credits
Manus (300/day), Genspark (100/day) and Tasklet (600 bonus credits on each day a paid user logs in) make returning daily worth something. Genspark's lifetime cap and Manus's Lite-only daily credits stop the free refill from becoming a permanent free tier.
Email isn't the hour-one lever, and identity walls cost sign-ups
In the first 1–2 hours, no product sent anything beyond a verification code or link; all the conversion pressure was in-app. Seven of eleven blocked a throwaway sign-up (work-email checks, Slack- or Google-only sign-up, a captcha, a PIN that never arrived, an instant suspension); five of those seven went through as soon as a real Google account was used. Manus, which suspended the throwaway account instantly, accepted the real account without question.
Standout tactics
- Native: paywall your own content. '47+ more posts, made for Kite. Pick a plan to unlock them' over a fan of the user's real posts, blurred and locked, with yearly preselected.
- Manus: one secret premium answer. The first task on the free 'Lite' plan quietly ran on the paid model, then a banner said 'You just got a free Manus 2.0 trial for this message! Upgrade to continue receiving the most refined, high-quality responses.'
- Viktor: out-of-credit messages escalate to the person who pays. The agent pauses in Slack and offers to 'Ask an admin'. The admin is told how many teammates are waiting, and paused requests replay once credits return.
- Tasklet: credits for activation. A 'Get set up' checklist pays +250 for connecting an app, +500 for automating a task, +1,000 for texting or sharing the agent; the free grant grows only when users do the sticky things.
- Lindy: personalization from the email domain. Before any setup, the home page suggested weekly research on the company's real competitors and prospect lists, inferred from the work email alone.
- Okara: the agent's log doubles as an upsell channel. A terminal-style log narrates real research on your company, then prints '[LinkedIn] LinkedIn posts unlocks on the paid plan'. Locked agents offer 'Ask the CMO what it would do'.
- Town: a reverse trial with a credit cap. Everyone gets Pro for 7 days or until the trial credits run out. Afterwards Free remains, but the automations built during the trial pause until you upgrade.
- Zapier Agents: a depth cap the agent itself announces. 'I've reached the limit of 10 activities per run on Agents' Free plan. Upgrade… to unlock up to 40 continuous activities per run. Shall I continue with the rest for now?' The free tier feels shallow on exactly the complex tasks people care about, and free users can still finish.
- Sintra: put the price on the button. 'Generate video · 51' after the free credits went on preparing it, plus an in-app picker that sells '+100 credits on us' and outcome counts instead of raw credits.
- Tasklet and Manus: daily credits. Paid users earn credits for showing up each day (Tasklet); free users get a daily trickle on a cheaper model (Manus).
Recommendations for Kite's trial-to-paid flow
Show finished work for the user's own brand in minute one, from the website alone
Ask for the URL, start the research immediately, and ask the 2–3 remaining questions while it runs (Native and Okara both do this). Land the user on 3 real deliverables (e.g. 3 LinkedIn posts plus a one-page competitor brief) before any credits are mentioned. For the many sign-ups with no website, offer a fallback in the same box (a Google Business Profile, an Instagram handle or two sentences) rather than a separate step.
AM-127: 72 of 219 sign-ups left at Kite's website step, and only 8 of the 18 who reached the $20 wall had a value moment first. Native: posts ready about 60 s after the URL. Lindy tailored its home suggestions to the company from the work-email domain alone.
Don't bill onboarding; pay credits for activation instead
Native's free unit is '3 posts', Okara's is 'your strategy + brief', and neither meters the setup work. Kite could grant an outcome bundle (say, 3 posts and 1 research brief), never charge for kickoff, onboarding or design work the user didn't ask for, and extend the AM-106 ladder with product-action rewards the way Tasklet does (+credits for connecting a channel, scheduling a recurring task, inviting a teammate).
AM-127: a median $9 of the first $20 went on unrequested onboarding, kickoff and design work. Tasklet's 'Get set up' milestones pay +250 to +1,000 credits each.
Put the paywall on the next batch of the user's own work
When the free unit is spent, show the already-generated next batch (the rest of the 30-day calendar, more posts, a deeper report) blurred and labelled for their brand ('22 more posts made for Acme are ready'), with one primary 'Unlock' button. A paywall that sells their own work converts better than one selling 'more credits'.
Native's paywall (steps 27–30), Okara's locked agents over blurred example work (step 15).
Translate credits into outcomes on the meter, the plan picker and expensive buttons
Show '≈ 14 posts or 3 research reports left' next to the dollar balance. Price big actions on the button before they run ('Generate video · $4.10'). Give each AM-106 ladder step an outcome-framed reward ('verify your phone: +$30 ≈ 15 more posts').
AM-127: nobody bought credits. Sintra, Native, Lindy and Viktor all explain credits as tasks or posts.
Let Kite deliver its own usage warnings, and route them to whoever pays
In Slack and web threads, Kite should warn at 70% and 90% with a run-out estimate, pause gracefully at zero ('I've paused this; want me to ask <admin>?') and replay the paused work after the upgrade. Teammates who hit the limit should generate an ask to the payer, not a dead end.
Viktor's Slack flow (steps 14–16); Zapier's in-run cap message (step 19); Town's and Gumloop's 75–90% emails. Sintra's toast-only 'out of credits' was a missed moment.
Narrate the work and tell the user what a paid plan would do next
Show a live activity feed while Kite researches and writes. Use the same voice for the next step that needs a plan: 'I can publish these every Tuesday and track replies; that's on Growth.' Offer 'ask Kite what it would do' on locked actions as a free preview. Manus goes further: the first task quietly runs on the premium model, then it tells the user and asks them to upgrade to keep that quality. Kite could run the first deliverable at full effort and say so.
Okara's Agent Log (steps 12, 16, 18) and the 'Ask the CMO what it would do' button (step 15). Manus post-task banner (Manus step 13).
Keep the trial no-card, but bound it by time and credits, and default to annual with a human touch
Town's reverse trial (7 days or the credits, whichever comes first, then a free floor) is the closest to Kite's ladder. Pair it with an annual default whose extra is human (Native's yearly-only '1-on-1 setup and strategy call'). State the renewal price on the card, as Native does, not in checkout fine print.
Native's yearly-only perks (step 37), Town's trial rules ([DOCS] steps 13–14), Sintra's renewal at $624 (step 3).
Re-engage at the paywall by email
None of the 11 sent a welcome or nurture email in the first hours. All the pressure was in-app, so a user who walks away at the paywall hears nothing. Send a 'your 22 posts for Acme are waiting' email a few hours after an abandoned paywall, with the blurred posts embedded.
Only verification emails arrived during the run: inboxes last checked 21:34 UTC, 1.5 h after the earliest sign-up (Native). Day-1 to day-7 emails are tracked in follow-up AM-138.
Blocked sign-ups Nikhil could unblock
- Viktor: needs an empty Slack workspace. Creating one with the Google login hit Google's 'Verify it's you' phone check; if you complete it in the live browser, I'll install Viktor only there.
- Town: signing up *is* granting inbox read/send and calendar edit for the Google account. Reply 'ok Town' to allow that on the test account; otherwise it stays documented from docs.
- Each Composio cloud browser session lasts roughly 20–60 minutes and drops the Google login when it ends, so each new session needs a fresh sign-in (and sometimes the phone check).
How this was captured
- 28 Oct 2026. Throwaway runs: a private headful Chrome (1440×900) on an India-hosted server with mail.tm/maildrop inboxes. Signed-in runs: a Composio (browser-use) cloud Chrome that Nikhil signed into with a real Google Workspace account; we used 'Sign in with Google' with identity scopes only and masked the name and email on every screenshot.
- Throwaway inboxes on mail.tm (maxxspace.com) and maildrop.cc; website entered was kite.ai and test tasks were 'research 3 competitors of kite.ai and draft a LinkedIn post (don't post)', so outputs show what each product would do for Kite.
- Never entered card details, connected real accounts, or let a product post, email or invite anyone. Checkout pages were screenshotted and abandoned.
- Steps marked docs come from the product's public help pages, used where sign-up was blocked.
- Personal data of other people (Native's public feed of other users' posts) is blurred. Marketing testimonials shown on the vendors' own sites are left as published.