Pricing

Two things decide your plan.

How many answers are included, and which model tiers you can use. Everything else is a limit listed further down, not a reason to move up. Two axes is what anyone can hold in their head, and we would rather sell you the right rung than a feature matrix.

Ask plans, compared by included volume, model tiers and plan limits
Plan Free
Free
Starter
$19 a month
Growth
$49 a month
Pro
$99 a month
Business
$249 a month
Above that
Talk to us
Answers included, in credits 200 a month2,000 a month5,000 a month10,000 a month25,000 a month Whatever you need
Model tiers you can use BasicBasic and mediumBasic, medium and advancedBasic, medium and advancedBasic, medium and advanced All three
Sign up Start free Start free trial Start free trial Start free trial Start free trial Contact us

The two shaded rows are what you are choosing between. The rest is reference.

Plan limits, for reference
Per-plan limits
Limit FreeStarterGrowthProBusiness
Bots 1251050
Widgets 1251050
Verified domains 131025100
Entries 2001,0005,00020,000100,000
Sources 131025100
Pages per crawl 502005001,0002,000
Shortest sync interval 7 days1 day1 day6 hours1 hour
Connected APIs --2520
Channels 11248
Replay credits 50 a month200 a month500 a month1,000 a month2,500 a month
Daily spend cap 13 credits133 credits333 credits666 credits1,666 credits
Webhook routing ---YesYes
Ask badge removed ---YesYes
Custom domain ---YesYes
Staff accounts UnlimitedUnlimitedUnlimitedUnlimitedUnlimited

The free plan is limited by capacity, not by capability

Every feature works on the free plan: states, cases, the inbox, email routing, replay, versions, analytics. 200 credits a month, the basic model, and an Ask badge on the widget that cannot be removed.

That is deliberate. A free user who never gets notified never opens the inbox, and a free tier that hides the inbox behind a paywall has just demonstrated a commodity chatbot. You should upgrade because you ran out of volume, not to unlock a screen.

One setup step that belongs only here: outbound email needs a verified domain. Cases appear in your inbox immediately either way; email routing starts working once your widget's domain is verified.

You are billed per answer

Not per conversation, not per visitor message. If somebody fires off three short lines in a row and the bot replies once, that is one answer. A long conversation where the bot answers eight times is eight.

Behind-the-scenes work is absorbed. An answer that had to search your content and call your API three times is still one answer, and you will not find a line item for retrieval or tool calls anywhere.

Model tiers are answer weights, not a second meter

One meter, in credits. What changes is how many credits an answer costs, depending on which tier produced it:

1 credit per answer

Basic

The default, and where most questions belong. Fast, and cheap enough that you stop thinking about it.

3 credits per answer

Medium

For questions that need to hold several things at once - comparing options, reading a long policy back to somebody.

16 credits per answer

Advanced

The expensive one, and priced honestly at that. Worth it where a wrong answer costs you a customer; not worth it for opening hours.

In practice: leave routine questions on basic and reserve the top tier for the ones that actually need it. A per-plan estimator is coming; for now, 200 basic answers is what the free plan gets you.

When the usage is gone, the bot stops

This is a feature, not a failure mode. There is no overage and no surprise bill. Warnings arrive at 80% and 95%; at 100% the bot tells visitors it cannot answer right now, their questions are still recorded as cases, and a top-up restores service immediately. Top-ups roll over.

What we never charge for

Questions people actually ask

What happens when I hit the limit?

The bot stops answering and says so. Visitors' questions still land in your inbox as cases, so you can see exactly who you turned away. Top up and it resumes at once.

What happens if I downgrade?

The change applies from the next period, and the credits you have already been granted for the period in progress stay yours. Nothing is deleted for being over a limit on the smaller plan; you just cannot add more of it.

Do you train on my data?

We do not build models, so there is nothing here for us to train. Answers are produced by third-party providers reached through an aggregation service, and what those providers do is governed by their own terms, which is why we publish the full set rather than making a blanket claim on their behalf. The sub-processor annex lists every one, with purpose and location.

Where is my data processed?

Your workspace's own data sits with us. Answers are produced by the provider set in the annex, and the aggregation service picks per request, so the honest answer today is "one of the providers on that list". Pinned regional processing is being built and will become a plan feature when it is verified end to end - not before, and you will not see the claim here until then. The longer version is on the security page.

Can I export everything?

Yes - the whole workspace, as a downloadable file, whenever you like. Deletion is separate and reaches the derived data too.

Do I need a card to start?

No. The free plan needs an email address and nothing else.