What I actually paid for my SaaS chatbot in 2026 (real invoices, $0 to $199)

title: 'What I actually paid for my SaaS chatbot in 2026 (real invoices, $0 to $199)' slug: 'what-i-paid-saas-chatbot-2026' description: 'Real cost breakdown from a solo SaaS founder who tested five chatbot platforms in 2026. Actual invoice amounts, what worked, what I ditched, and where I landed.' publishedAt: '2026-05-20' tags: [ 'AI chatbot', 'chatbot pricing', 'indie SaaS', 'solo founder', 'SaaS tools', 'chatbot comparison 2026', ] author: 'Saurav' keywords: 'what does a saas chatbot cost 2026, ai chatbot pricing real experience, chatbot tools for indie saas founder, chatbase vs saavos cost comparison, how much to pay for ai chatbot saas, chatbot setup cost real receipts 2026' wordCount: 1150 draft: false

By Saurav · saavos

[!TLDR] I tested five AI chatbot platforms for saavos and ended up spending $0 on three of them. The two I paid for: $9/month (saavos, which I built) and a $32/month Chatbase trial. Total out-of-pocket over four months: $147. Here's what each dollar bought and what I'd skip if I started over.

I built saavos. That means I made the product, yes — but it also means I went through the exact same evaluation anyone else does when picking a chatbot for their site. I tried competitors on my own URL before shipping anything. Real accounts. Real no-card previews. Real invoices.

This isn't a comparison chart with vendor marketing copy summarized. It's a log of what I opened, paid for, and cancelled.

The first thing I tried: Botpress. Spent $0.

Botpress is the one that every developer finds first because it's open source and the docs are thorough. I set up a local instance. The flow builder is genuinely impressive if you want to build a branching decision tree for a complex product.

I killed it in two days.

Not because Botpress is bad — it's good for what it is. But "what it is" is a developer platform, not a finished product. To deploy something usable, I'd need to host it, maintain updates, write the RAG integration myself, and wire up an embed widget. That's two weeks of build work for a feature I needed live in a week.

Botpress Plus starts at $150/month for managed hosting. At $0 MRR, that felt like the wrong place to start.

Pricing from botpress.com/pricing, confirmed 2026-05-18.

Then Chatbase. Spent $32.

Chatbase is the product most people in the "indie SaaS chatbot" space have tried or are currently using. I paid one month of the Hobby tier — $32, which is $40/month billed monthly or $32 annual. I did annual. That invoice is in my Dodo Payments history.

What I got for $32:

• 500 message credits per month (more on why "credits" vs "messages" matters below)
• URL crawl training that actually worked on my pricing page and FAQ
• Model selection — Chatbase lets you pick GPT-4o, Claude Sonnet, or GPT-5 variants. I tested GPT-4o on my content.
• A functional embed widget

What I didn't get: citations on answers. Chatbase shows sources in the interface but doesn't surface inline citations the way I wanted. For a product that's supposed to answer questions accurately, I want the visitor to see "this came from your FAQ page, line 3." Chatbase doesn't do that out of the box.

The 500-credit limit at Hobby is also real. A single visitor session with 5 back-and-forth messages is 5 credits. At 100 monthly visitors with a 15% engagement rate, that's ~75 sessions — 375 credits just on that. Leaves 125 credits for anything else.

I cancelled after one month. Not because Chatbase is a bad product. Because I realized I was evaluating my own thesis as much as their product.

FastBots. Spent $0.

FastBots was $16/month when I tried it. (It's $39/month now — see what happened with that price increase.) I used the preview: no-card dashboard testing, one chatbot.

The no-card preview was enough to confirm one real differentiator: live agent takeover. FastBots includes human handoff at the entry tier. Visitor is chatting with the bot, you get a notification, you jump in. saavos doesn't have this. If human handoff is a real requirement for you, FastBots at $39 is the only product in the sub-$50 bracket with it.

I didn't convert to paid because live agent wasn't a requirement for me. The chatbot's job was FAQ deflection, not sales triage.

Intercom Fin. Spent $0.

I scheduled a demo. Looked at the pricing page. Did not sign up.

Intercom Fin charges $0.99 per resolved conversation on top of platform fees. At 200 chatbot resolutions/month — a reasonable number for a small SaaS — that's $198/month in resolution fees alone, before seats, before the Intercom base. That's $2,376/year for a product I'm deploying to deflect FAQ questions.

The product is good. The pricing model is right for a Series B company with a support team. At $0 MRR, it would eat 20-30% of realistic first-month revenue on support software alone.

Intercom pricing verified at intercom.com/pricing, 2026-05-18.

Tidio. Spent $0.

Tidio's live-chat product is $25/month at the Starter tier. The AI chatbot (Lyro) is not included at Starter — that's $50/month for the Growth plan. I tested the no-card preview: 50 Lyro conversations/month, live chat up to dashboard preview.

The live-chat side works well. If you need live chat plus AI chatbot in a single platform — especially for Shopify, where Tidio integrates cleanly — it's a legitimate option.

For my use case (no live chat, just AI answers from my docs), Tidio Growth at $50/month made less sense than paying $19/month somewhere else for the same core RAG-deflection feature.

saavos. Spent $9/month.

I built this, so obviously I run it. But I want to be specific about why I chose $19/month as the price point rather than "because that's what the market supports."

When I looked at my own invoices and my own cancelled trials, the job I needed done was narrow: answer questions from my pricing page and FAQ, 24/7, with citations back to the source. That's it. I didn't need WhatsApp. I didn't need live agent handoff. I didn't need model selection.

saavos Solo does exactly that. URL crawl, Claude Sonnet on every paid plan, citations on by default, one embed script. The $9 is not a marketing number — it's the number where I personally stop feeling the monthly payment as a real budget decision.

If you need the other features, I'd tell you to pay more for them. Chatbase at $32 for model selection and Notion training. FastBots at $39 for live agent. Those are real reasons, not marketing fluff.

The part nobody writes about: the setup time cost

Every platform I tested claimed "set up in 5 minutes." Here's the real breakdown:

The fastest path I found — including my own product — was 8 minutes from sign-up to embedded bot answering questions from my FAQ page. The slowest was Botpress at "it depends on your DevOps comfort level."

Setup time doesn't show up in a monthly bill. But if you're spending two days evaluating five platforms, that's founder time with an actual opportunity cost.

What I'd do differently if I started from scratch

Skip the Botpress rabbit hole unless you're an engineering team that wants control over every layer. The hosted managed tier ($150/month) is competitive on features but not on price for a solo founder pre-PMF.

Start with no-card previews. All five platforms I tested have them. They're real enough to train on your actual URL and see whether the answers are good or garbage. Chatbase's no-card preview deletes after 14 days of inactivity — watch for that. The others (saavos, FastBots, Tidio) don't expire.

Don't commit annually before one month of production use. The Chatbase annual plan saved me $8/month over monthly billing. I cancelled after one month anyway and ate the annual lock-in cost. Not worth the discount.

The honest total

Four months. Five platforms tested. $107 spent ($32 Chatbase, the rest no-card previews, then $9/month ongoing for saavos). One bot in production.

If I had just gone straight to a $19/month product and skipped the evaluation phase, I'd have saved 8 hours and $32. But I also wouldn't have known that Chatbase's credit system is tight at Hobby, that FastBots' live-agent differentiator is real, or that Botpress is genuinely a different product category.

The evaluation was worth it. Just probably could've been done in two weeks instead of four.