On February 2, 2023, a university student named Yasser Elsaid tweeted a link to a weekend project out to his 16 followers. He called it Chatbase, and the whole pitch was "ChatGPT for your documents": upload a PDF, then chat with it. His first paying customer showed up within 30 minutes.
Three years later, that side project is a bootstrapped company pulling in $10 million a year in revenue, with zero outside investors and paying customers spread across 80 countries. Chatbase is a no-code platform that turns your own content, a website, a PDF, a help center, into a custom AI chatbot you can embed on your site, connect to WhatsApp and Slack, or hand real phone calls to. No engineer required.
Here is the question this review actually answers: is Chatbase still the cheap, scrappy chatbot builder that made it famous on Product Hunt and Reddit, or has it quietly repositioned itself into an enterprise AI agent platform that costs a lot more than you remember? Short answer: a bit of both, and which one you get depends entirely on which pricing tier you land on.
Quick Verdict
Metric | Detail |
|---|---|
AIwithKay score | 7.5/10 |
Best for | Support and sales teams who want a custom AI agent trained on their own content, live within a day |
Starting price | Free (50 message credits/month); paid plans from $40/month, $32/month billed annually |
Free plan | Yes, but inactive agents are deleted after 14 days of no use |
Bottom line | A genuinely fast way to turn your own content into a working chatbot, just budget carefully around the credit system, since real traffic burns through it faster than the sticker price suggests |
What Works Well / What To Watch
What works well:
Setup really is fast. Point Chatbase at your website, let its crawler pull your pages, or upload a stack of PDFs, and you have a working agent the same day, no code involved.
Model choice, not model lock-in. You can pick which AI model answers your questions, from lighter, cheaper options to heavier frontier models, rather than being stuck with whatever the vendor decided was the default.
Voice and telephony, not just a chat widget. Standard plan and up can answer actual phone calls, something most website-chatbot builders at this price simply do not offer.
A genuinely usable free plan. 50 message credits a month costs nothing and is enough to kick the tyres before you commit a card.
Real integrations, not just a webhook and a prayer. Native connectors to Shopify, Stripe, Salesforce, Zendesk, Calendly, and Slack mean the bot can actually do things, like look up an order or book a call, not just answer questions about them.
What to watch:
The credit system punishes success. Think of it like a mobile data plan bought back before unlimited plans existed: cheap on paper, until your bot goes viral and you are buying emergency top-up credits at a worse rate than your base plan.
The jump between tiers is steep. Hobby to Standard nearly quadruples your bill for 8x the message credits, an awkward middle for a team that has outgrown the cheap plan but is not ready for the expensive one.
Removing the badge costs extra. The "Powered by Chatbase" tag stays on your widget unless you pay an additional $99 a month for white-labeling, even on plans that already cost hundreds.
The reviews do not agree with each other. Chatbase sits at 4.8 on G2 from a small pool of 19 reviews, but at 3.7 on Trustpilot from a larger, angrier crowd, where billing and refund complaints are the recurring theme rather than the bot's actual answers.
Training data caps are tight lower down the stack. 1 MB on the free plan and 10 MB on Hobby is not much if your help center or product docs run long, and you will hit that ceiling before you hit the message-credit ceiling.
Who Built Chatbase (And Why the Backstory Matters)
Elsaid was in his final year of university when the PDF-chat tool he had built in about two months went viral off that one tweet. Six months later he was at $64,000 in monthly recurring revenue. Two years in, he had crossed $5 million in annual revenue. As of mid-2026, Chatbase is doing more than $10 million a year, still with no venture funding, still run by the same founder, now with a team of roughly 30 people instead of one.
That growth story matters because it explains the product's identity shift, in a good way. Chatbase started as a simple document-chat toy. It has since rebuilt itself around what its own homepage calls "the leading AI agent for CX," with dedicated Support, Sales, and Product Guidance agent modes, omnichannel deployment, and analytics that track resolution rates and customer sentiment. Its customer list, per its own site, now includes Dolby, Pearson, Roche, Bridgestone, Sage, and yes, Chuck E. Cheese, a genuinely odd but real mix that says less about any one industry and more about how broadly "answer customer questions automatically" applies.
That repositioning puts Chatbase in the same conversation as enterprise AI agent platforms like Sierra, Decagon, and Fin, the company formerly known as Intercom, which we have reviewed here. The difference, and it is the honest reason to still consider Chatbase over those names, is that Chatbase kept a self-serve, published price tag. Sierra and Decagon make you talk to sales before you learn what anything costs. Chatbase still lets you sign up with a credit card at $40 a month.
Key Features
Custom training on your own content. Upload PDFs, DOCX, TXT, or CSV files, paste text directly, crawl your whole website automatically, or sync a Notion workspace, and the agent learns from all of it.
Retrieval-augmented answers. Rather than trying to memorize your documentation, the model works more like an open-book exam: it looks up the relevant passage from your own content before answering, which cuts down on made-up answers compared to a bot working from general knowledge alone.
Multi-channel deployment from one agent. The same trained bot can sit on your website as a widget, answer in WhatsApp, Slack, Telegram, or Messenger, and handle inbound phone calls on Standard plans and up.
AI Actions. Native connectors let the bot look up a Shopify order, process a Stripe billing question, update a Salesforce record, or book a Calendly slot, real actions, not just canned replies.
Helpdesk handoff. When the bot does not know the answer, or a customer explicitly wants a human, the conversation routes into your existing helpdesk queue with full context attached.
Auto-retraining. On Standard plans and above, the agent automatically re-crawls and re-ingests your source content on a schedule, so it does not quietly go stale as your docs change.
Analytics that actually tell you something. Conversation volume, resolution rate, and a log of what customers are actually asking, useful for spotting a documentation gap before it becomes a pattern of complaints.
How Chatbase Actually Works, Step by Step
Worth saying plainly before this section: we have not tested a paid Chatbase account ourselves. What follows is drawn directly from Chatbase's own published onboarding documentation and product pages, not a first-hand trial.
According to Chatbase's own "Build Your First AI Agent" guide, setup starts from an Agents dashboard, where you click "New AI agent" and choose how to train it. You can crawl a website (Chatbase's crawler discovers and indexes every linked page it can reach, then shows you the total character count it pulled in for training), upload files directly, paste text by hand, or connect a Notion workspace for automatic syncing.
Once the source content is in, you pick which AI model answers customer questions, from lighter default models to pricier frontier options, since heavier models cost more message credits per reply than lighter ones. From there, the agent gets an embeddable widget for your site and, depending on plan, can be wired into WhatsApp, Slack, Zendesk, Salesforce, or a phone number for voice support. Chatbase's own documentation frames the whole process as achievable without a developer, and multiple independent write-ups from actual users describe the same thing: point it at your content, and you have something answering questions the same day.
Chatbase Pricing in 2026: What You Actually Pay
Verified directly on Chatbase's own pricing page today.
Plan | Price | What's Included |
|---|---|---|
Free | $0/month | 50 message credits/month, 1 AI agent, 1 MB training content, 1 seat. Inactive agents are deleted after 14 days. |
Hobby | $40/month ($32/month billed annually) | 700 message credits/month, 10 MB training content, 2 seats, advanced models, integrations, basic analytics, attachments |
Standard | $150/month ($120/month billed annually) | 4,000 message credits/month, 20 MB training content, 3 seats, helpdesk handoff, voice, telephony, outbound campaigns, API access, auto-retrain |
Pro | $500/month ($400/month billed annually) | 15,000 message credits/month, 40 MB training content, 5 seats, advanced analytics, source suggestions, tickets as a training source |
Enterprise | Custom | Higher limits, SSO, white-labeling, audit logs, HIPAA-eligible, zero data retention, dedicated success manager |
Add-ons, on top of any plan: auto-recharge credits at $40 per 1,000, extra AI agents at $25 per agent per month, and removing the "Powered by Chatbase" badge at $99 a month.
Is it worth the price? For a small team testing the waters, the free plan and the $32-a-month Hobby tier are genuinely fair value, cheaper than hiring even a few hours of part-time support staff. Where it gets less comfortable is the jump to Standard, $120 a month for the plan that actually unlocks voice, telephony, and helpdesk handoff, features a lot of buyers will need on day one, not as an eventual upgrade. Budget for that jump before you fall in love with the Hobby-tier price.
Chatbase vs Intercom Fin vs Zendesk AI vs Tidio vs Voiceflow
Chatbase's honest positioning is "a very good agent builder, not a full support platform." Here is how it stacks up against the tools people actually compare it to.
Tool | Best For | Starting Price | Key Differentiator |
|---|---|---|---|
Chatbase | Teams that want a fast, self-serve custom AI agent without a sales call | $40/month ($32 annually) | Published, self-serve pricing even as it repositions upmarket |
Fin (Intercom) | Teams wanting transparent outcome-based pricing at enterprise scale | $0.99 per resolution plus seat fees | Pay only for resolved conversations, not a flat seat fee |
Zendesk AI | Teams already running Zendesk's helpdesk | $19 to $115 per agent/month, plus per-resolution AI fees | AI is native to a support suite most agents already use |
Tidio | E-commerce stores needing live chat plus AI in one inbox | Free; Growth plan $59/month, Lyro AI add-on separate | Human agents and AI bot share the same live-chat inbox |
Voiceflow | Teams designing complex, multi-step conversation flows, including voice | Free; Pro from $60/month, credit-based | Visual, no-code canvas built for complex conversation logic |
Fin and Zendesk AI both price by the resolved conversation, not the message, which sounds friendlier until your volume is high enough that outcome-based pricing gets expensive fast. Our Fin review and Zendesk AI review both dig into that math in more detail.
Tidio makes more sense if you want a human agent and an AI bot working the same inbox rather than a standalone widget. Voiceflow is the right call if you need a visual canvas for genuinely complicated, multi-step conversation logic, including voice assistants, rather than a bot that just answers questions from your docs.
Who Should Use Chatbase (And Who Should Skip It)
Best for:
Support or sales teams that want a working AI agent trained on their own content live within a day, without hiring a developer.
Businesses that need the bot to take real phone calls, not just handle a website chat widget.
Anyone who wants to compare AI models on the same content rather than being locked into one vendor's model.
Skip it if:
Your traffic is high enough that a credit-based pricing model will cost more, and less predictably, than a flat seat fee.
You need deep customization or advanced reporting; multiple users note Chatbase's analytics and configuration options are thinner than dedicated CX platforms.
You already run Zendesk or Salesforce at scale and would rather keep AI native to that platform than bolt on a separate tool.
Final Verdict
Chatbase earns its reputation the practical way: it is genuinely fast to set up, it does not lock you into one AI model, and it is one of the few tools in this category still letting you sign up with a credit card instead of sitting through a sales call. The founder-built, no-VC-money story is not just a nice detail either, it is a reasonable signal that the company answers to its customers' bills, not a board's growth targets.
The catch is the same one that shows up in almost every credit-based SaaS tool: the price that gets you in the door is not the price you will pay once you actually use the thing. Budget for the jump to Standard if you need voice or helpdesk handoff, and keep an eye on your message credits the way you would watch a phone bill on an old-school data plan.
If you run a support or sales operation and want to see what your own content looks like as a chatbot, Chatbase's free plan costs nothing to test. Just read the pricing table above before you pick a paid tier, not after your first overage bill.
GOT ANY QUESTIONS LEFT?
Is Chatbase free to use?
How much does Chatbase actually cost each month?
What's the difference between Chatbase and just using ChatGPT?
Does Chatbase have an affiliate program?


