Included
Free competitor teardown
We test the assistant you already run, on your live site, with your questions. Nine scored rows and the raw answers. We benchmark your current assistant before asking you to switch.
Pricing
A small site should cost almost nothing to run an assistant on. These are the plans we intend to launch with. Creobot opens to teams in order, and there is no checkout on this page.
Every AI reply uses credits, and how many depends on the model the agent runs on. The default, GPT-4o Mini, costs 1 credit a reply, so Starter’s 2,000 credits is about 2,000 replies. GPT-4.1 Mini, GPT-5 Mini and Gemini 2.5 Flash cost 2, Claude Haiku 4.5 costs 4, and the largest models cost more, up to 20 for Claude Opus 4.7. The model is chosen per agent, so you decide the trade between cost and depth.
A source is one website Creobot crawls. Plans cap both the number of websites and the pages crawled on each: 10 on Free, 100 on Starter, 500 on Pro and 1,000 on Scale. More pages do not automatically mean better answers, so pick the sections visitors actually ask about.
Storage covers uploaded files rather than crawled pages. A typical PDF handbook is well under a megabyte, so storage is rarely the binding limit. If it becomes one, that usually means the wrong files are being uploaded.
An agent is one configured assistant with its own sources, tone and rules. One site normally needs one. Agencies running a separate assistant per client site are the reason the higher tiers carry more.
Team members are the people who can sign in to the workspace: 1 on Free and Starter, 3 on Pro and 10 on Scale. More can be added as an add-on.
Custom credit limits, unlimited agents, websites, pages, storage and team members, SSO and SAML, dedicated infrastructure, audit logs and a priority support SLA. Priced with you.
Everything each plan includes, side by side. Anything marked planned is not built yet and is named that way deliberately.
| Feature | Free | Starter | Pro Most picked | Scale | Enterprise |
|---|---|---|---|---|---|
| Monthly | $0 | $29 | $99 | $199 | Custom |
| Usage | |||||
| Credits per month | 100 | 2,000 | 7,500 | 20,000 | Custom |
| Agents | 1 | 1 | 3 | 5 | Unlimited |
| Websites | 1 | 5 | 10 | Unlimited | Unlimited |
| Pages per website | 10 | 100 | 500 | 1,000 | Unlimited |
| File storage | 1 MB | 10 MB | 50 MB | 100 MB | Unlimited |
| Team members | 1 | 1 | 3 | 10 | Unlimited |
| Lead forms per agent | Not included | 1 | 3 | Unlimited | Unlimited |
| Deployment | |||||
| Playground testing | Included | Included | Included | Included | Included |
| Deploy widget to a live site | Not included | Included | Included | Included | Included |
| Hosted help page | Not included | Included | Included | Included | Included |
| Widget tabs: Home, Help, News | Not included | Not included | Included | Included | Included |
| Remove "Powered by Creobot" | Not included | Add-on | Add-on | Add-on | Add-on |
| AI models | |||||
| GPT-4o Mini, GPT-4.1 Mini, GPT-5 Mini | Included | Included | Included | Included | Included |
| Claude Haiku 4.5 | Not included | Included | Included | Included | Included |
| GPT-4o, GPT-4.1, GPT-5, o4 Mini, Claude Sonnet 4.6, Gemini 2.5 Flash | Not included | Not included | Included | Included | Included |
| Claude Opus 4.7, Gemini 2.5 Pro | Not included | Not included | Not included | Included | Included |
| Conversation | |||||
| Grounded answers with citation | Included | Included | Included | Included | Included |
| Refusal when a source does not cover it | Included | Included | Included | Included | Included |
| AI chat summaries | Not included | Not included | Included | Included | Included |
| Live human handoff | Not included | Not included | Not included | Included | Included |
| Analytics | |||||
| Conversation log | Included | Included | Included | Included | Included |
| Analytics and content gaps | Included | Included | Included | Included | Included |
| Integrations | |||||
| Slack, Cal.com, Calendly, Webhooks | Not included | Included | Included | Included | Included |
| Make | Not included | Not included | Included | Included | Included |
| Zapier, HubSpot, WhatsApp | Not included | Planned | Planned | Planned | Planned |
| Security and support | |||||
| Documentation and glossary | Included | Included | Included | Included | Included |
| Email support | Not included | Included | Included | Included | Included |
| Priority support | Not included | Not included | Included | Included | Included |
| SSO and SAML | Not included | Not included | Not included | Not included | Included |
| Priority support SLA | Not included | Not included | Not included | Not included | Included |
Scroll the table sideways to compare all five plans.
Before you pay anything
Already running Chatbase, SiteGPT or Intercom? Keep it running. Everything below happens before there is anything to buy.
Included
We test the assistant you already run, on your live site, with your questions. Nine scored rows and the raw answers. We benchmark your current assistant before asking you to switch.
Included for pilots
We map sources, configure Creobot and run side-by-side QA. Source inventory is the week nobody has scheduled, so it is the part we do rather than the part we ask you for.
Available now
Test on public pages, docs and approved sources without sending sensitive data. No customer records, no PHI, no payment data, no credentials.
Talk to us
For teams that need tighter control over pilot scope, accounts and access. This is a narrower pilot shape, not a certification, and it does not make Creobot compliant.
Included for pilots
Security questionnaire, data-flow summary, pilot rules, deletion process and subprocessor status. Sixteen questions answered directly, including the ones where the answer is no.
WHY NONE OF THIS IS PRICED
Creobot opens to teams in order and there is no checkout. Putting a number on migration work before the product can be bought would be inventing one. These are included for pilots because a pilot is what exists today.
The assistant stops replying until credits reset on your billing date. An extra credit pack or a plan upgrade applies straight away.
New websites are refused until you remove one or move up a plan, and pages past the plan’s per-website cap are not crawled. Existing answers are unaffected.
Uploads are refused. Crawled pages do not consume storage, so a site can keep growing without hitting this.
Credits reset every month on your billing date. Unused credits do not roll over.
Available on any paid plan and billed with it. The last two are named so the table is not quietly hiding them.
$15/mo per 1,000 credits a month
$150 a year on yearly plans. Adds to your monthly allowance straight away and at every reset.
$9/mo per agent
$90 a year on yearly plans. For agencies running one assistant per client site.
$19/mo per workspace
$190 a year on yearly plans. Takes the badge off the chat widget.
$5/mo per member
$50 a year on yearly plans. For teams bigger than the plan includes.
Not offered yet
Choosing a processing region. Asked for often, not built, and not promised.
Not offered yet
Planned before paid rollout rather than sold as an upgrade.
There is no trial to start and no card to enter. A pilot is scoped and priced with you first. That removes the usual risk question and replaces it with a different one: what happens after launch if you pay and it does not fit.
The positions we intend to launch with, stated now so they can be held against us. Monthly billing by default, so leaving costs you at most one month. Upgrades apply straight away and are charged pro rata; cancelling keeps your plan until the period you paid for ends, with no penalty. Your content and your conversation log belong to you.
What we are not offering is a money back guarantee. Saying so is more useful than a guarantee badge we would have to withdraw.
The honest way to de-risk this before launch is to tell us what your site has to answer. If it is a poor fit we will say so, and that conversation costs nothing. The comparisons name the cases where another tool is the better choice.
Leaving costs you at most one month. No annual lock in to get a usable plan.
Upgrades are charged pro rata. No cancellation penalty and no downgrade fee.
Your content and conversation log belong to you.
Not offered. Cancelling keeps your plan until the period you paid for ends.
What the first pilot covers
The first pilot runs on public website content, docs, help pages and sources you approve, which is why it can start before a security review does. Stated plainly: Creobot holds no SOC 2, HIPAA, ISO 27001 or PCI DSS certification, and makes no zero-retention or no-training guarantee on behalf of its model providers. The subprocessor list is published; retention periods and a signed DPA land before anyone is charged. Read the security page for what is settled and what is not.
No. Creobot opens to teams in order and these are the prices we intend to launch with. They are a direction, not an offer, and they may move before self serve opens. Nothing on this page can be purchased today.
Credits pay for AI replies. Each reply uses between 1 and 20 credits depending on the model the agent runs on: 1 for the default GPT-4o Mini, 2 for GPT-4.1 Mini, GPT-5 Mini and Gemini 2.5 Flash, 4 for Claude Haiku 4.5, and more for the largest models. Retrieval is included, and if the first model is unavailable the fallback reply is charged at the model that actually answered.
One source is one website Creobot crawls. Each plan also caps the pages crawled per website: 10 on Free, 100 on Starter, 500 on Pro and 1,000 on Scale. Uploaded files count against storage, not sources.
The assistant stops replying until your credits reset on your monthly billing date. To keep going sooner, add an extra credit pack (1,000 credits a month for $15) or move up a plan; both apply straight away.
The Free plan is not time limited. It gives you the full playground, one agent and 100 credits a month for as long as you like, so you can judge the answers on your own content before paying. Putting the widget on a live site starts at Starter.
No. Credits reset monthly. Rolling credits over would make capacity planning harder on both sides, and we would rather the plan you are on reflect the traffic you actually have.
Yes. An upgrade applies immediately, with the difference charged pro rata, and your credit allowance moves up the same moment. A downgrade applies immediately without a refund. Cancelling keeps your plan until the end of the period you paid for, then moves the workspace to Free.
Ask. Pricing is not final, so a question about your actual traffic and source count is more useful to both of us than the table above.
Launch
Pricing is agreed with you before anything starts. Tell us your traffic and how many pages you would train on, and we will work through what it would come to.