Read Retool's pricing FAQ, not its homepage

There is a sentence sitting in Retool's pricing FAQ, below the fold, inside a collapsed accordion, that a competitor would pay for: "When you need to distinguish between them, the previous experience is referred to as the classic app builder."

That is a company telling you it rewrote its core product and now has two of them to support. It is not in the hero copy. It is not in a press release. It is in the eleventh question of a FAQ that exists to reduce sales friction.

Pricing FAQs are written by a different instinct than the rest of a marketing site. Homepages are written to attract. FAQs are written to preempt — every question on that list is a question that came up often enough in real sales conversations to be worth answering before a human has to. The page is a transcript of objections. Read enough of them and you're reading the company's own view of where it's weak.

Three seat types is an admission about who actually logs in

Retool's FAQ splits users into builders, internal users, and external users, and notes that the split is an Enterprise-plan capability. Builders create and publish. Internal users are employees who "can explore and draft freely without consuming a builder seat." External users are partners and clients.

Work backwards from that. You don't invent three seat classes because your pricing model is elegant. You invent them because large accounts pushed back on paying builder rates for people who open an app twice a week. The per-user rates published alongside it — $50 and $15 and $10 and $8 and $6 and $4 across tiers and user types — describe an organization that has been negotiated with, repeatedly, by procurement teams who counted seats.

The strategic read is not "Retool has flexible pricing." It's that the buyer who matters to them now has hundreds of occasional viewers and a handful of builders, which is a very different customer than the twelve-engineer startup the product was originally shaped around. That shift shows up in the seat table years before it shows up in anyone's positioning. It's the same move Datadog's price list makes with its per-product line items: the units of the price sheet are the units of the org.

Calling your own product "classic" is a dated event

The rename is the most useful line on the page, because renames have timestamps.

"Classic app builder" is a term that exists only after a rewrite ships and only while the old thing still has enough users to need a name. Both halves are informative. The rewrite shipped — that's a completed bet on an AI-native builder, not a roadmap item. And the old one persists, which means a migration is in progress rather than finished, and some portion of the install base is sitting on the previous experience for reasons nobody has solved yet.

If you compete with Retool, that window is the whole opportunity. Migration periods are when accounts are most reachable: workflows break, muscle memory breaks, and somebody has to re-approve a tool that suddenly feels unfamiliar. The FAQ even carries the reassurance — apps built in the previous builder continue to work — which is the sentence you write when customers have asked whether they will.

Hourly agent billing is a bet placed in public

The clearest strategic commitment on the page has nothing to do with apps. Retool bills AI agents by the hour, at rates that vary by model, measured as wall-clock time from task start to finish — "including executing steps, waiting on API responses, and model processing," excluding idle time waiting on a human.

Almost nobody prices agents this way. The default is tokens or runs. Retool's own FAQ makes the argument out loud: hourly "mirrors how teams already value labor: you pay for productive time, not tokens or runs."

That is a pricing thesis, and it is falsifiable. It says Retool believes it is selling labor replacement rather than compute, and that its buyers will accept a bill that looks like a contractor invoice. It also means their margin is exposed to model latency in a way a per-token competitor's isn't — a slow model run costs the customer more and buys Retool nothing extra. Watch for that unit to change. If hourly quietly becomes "hourly, with a run cap," the thesis met the spreadsheet.

Two more numbers on the same page complete the picture. Bring your own model key and Retool credits aren't consumed — the LLM cost is theirs to eat otherwise. And workflow overages run $75 per 5,000 runs. Those two coexist uneasily; one is a margin release valve, the other is a metered fallback.

The deadline in the footnote

Then there's this: customers who sign an Enterprise contract by September 30, 2026 receive AI credits worth up to $10k per year of the contract term.

A dated incentive with a hard cutoff a week out is a company pulling deals into a quarter. That's not a scandal — everyone does it. But it is a discountable, time-boxed, publicly-posted concession on Enterprise contracts, which is exactly the fact a competing rep wants in hand during a bake-off. It also sets a price on the credits themselves, which the rest of the page carefully avoids doing.

Credits get quantified elsewhere in the same FAQ: roughly 12 per prompt, 3–6 prompts per app, 36–84 apps per month on Business. That's an unusual amount of arithmetic to publish. It reads as a response to a specific objection — how do I know I won't run out? — which suggests enough prospects asked it to make the vague version untenable.

What this is actually worth

Most competitive reading is low-yield. Homepages change for reasons that have nothing to do with strategy, and blogs are written to be quoted. Pricing FAQs are neither: they're defensive documents, maintained reluctantly, updated when reality forces it. The signal density is higher than anything else on the site, and the radical-transparency companies that publish everything are, counterintuitively, harder to read than a company that publishes one honest page.

The catch is timing. A pricing FAQ's value is in the delta — the day "classic app builder" first appeared, the day hourly agent billing replaced whatever came before it, the day a dated discount went up. Read the page cold today and you get a snapshot without a clock on it.

That's the specific problem Seeto exists for: it watches public surfaces continuously and surfaces the changes as discrete, timestamped events, so the rename and the new billing unit and the deadline arrive as things that happened on a date rather than things you noticed eventually. It won't tell you what the hourly bet means — that read is yours, and it's the part worth your time. It will tell you the day the bet was placed.

Go open the pricing FAQ of your closest competitor right now. If it's three years stale, that's a finding too.