Crisp Pricing in 2026: The Flat Rate That Still Meters Your AI

Crisp's pricing page tells a great story: flat per-workspace plans, unlimited conversations, no per-seat surprises. It's the rare support tool that rejects per-conversation billing out loud. And then the fine print quietly adds a second product - Hugo, its AI agent - paid for with dollar-denominated credits that run out mid-month. The flat rate is for the humans. The AI is a meter with a very different shape.

Here's the honest 2026 breakdown of Crisp's pricing model: how the plans work, what the AI credits really cost, when the math stops being flat, and what to look for instead.

Part 1: The per-workspace plans

As of this writing, Crisp lists four self-serve plans, all priced per workspace per month:

	Free	Mini	Essentials	Plus
Price / month	$0	$45	$95	$295
Seats included	2	4	10	20+
AI credits included	$0	$5	$25	$75
Approx. AI conversations	-	~90	~450	~1,350

That much is genuinely refreshing. Unlike Intercom or Zendesk, Crisp does not bill per agent seat, and it doesn't cap your conversation volume on the human side. Extra agents on the plans where Crisp publishes a seat price run around $10/month each - cheap by industry standards. For a small team that just wants live chat plus a shared inbox, there is a lot to like.

The catch is the last column, and it's the one the headline doesn't mention: the AI is not part of the plan. The AI agent, Hugo, is a second billing surface that lives on credits.

Part 2: The AI credit meter nobody puts in the hero

Crisp denominates its AI in dollars of included credit, then annotates each allowance with an approximate conversation count - $5 for around 90 automated conversations, $25 for around 450, $75 for around 1,350. The credits are token-based: the longer the conversation, the more context loaded from your knowledge sources, and the more detailed your replies, the more credit a single conversation burns. Crisp's own estimate puts a full Hugo-handled conversation at roughly $0.05 to $0.10 of credit.

Read that again, because it's the whole story: an "automated conversation" is not a fixed unit. It's a budget that depends on message length, how many times you go back and forth, how big your knowledge base is, which AI model you select, and how many replies Hugo sends. The "around 90 conversations" is a model, not a promise.

A conversation cap that depends on how chatty your customers are is not a fixed allowance - it's a weather report.

When the credits run out, Hugo stops answering and the conversations escalate to your team. Crisp's default is to drop the bot at the end of the allowance unless you switch on pay-as-you-go and let the meter run past what you prepaid. So the experience of "AI that resolves tickets" quietly becomes "AI on Monday, humans by Wednesday" - then a second surprise when the top-up bill arrives.

Where the flat pricing stops being flat

Now put the two parts together and watch what happens at volume:

• A good bot increases your meter. The better Hugo gets, the more conversations it resolves - and each resolved conversation consumes credit. High deflection is exactly what drains the allowance, on top of the plan.
• Billable events multiply. A conversation that jumps between Hugo and a human still burns AI credit for the AI-handled portion, so the same ticket can cost twice - once in credit, once in agent time.
• The fixed price is a floor, not a ceiling. Plan cost is predictable; the AI line item is roughly a quarter of the monthly bill on the paid plans (about 26% on Essentials, 25% on Plus) and the conversation allowance is what breaks first under real traffic. It's a flat-rate product with a usage-based rider.
• Every plan is a different product. Hugo only starts on Mini, the omnichannel chatbot and knowledge base only fully arrive on Essentials, and the AI-first support suite is Plus territory. The "AI chatbot" you read about depends on which tier you're on.

None of this makes Crisp a bad product. It makes it a two-meter product wearing one flat-rate costume.

Crisp vs. the rest of the field

Crisp sits in an interesting middle of the pricing spectrum. Intercom and Zendesk add per-seat charges and per-resolution or per-agent AI fees on top of everything. Chatbase and CustomGPT scale their bills with document usage and model credits. Tidio bills Lyro per conversation on top of the inbox plan. Crisp is arguably the friendliest of that set for a small team - the per-workspace model, unlimited human conversations, and cheap extra seats are genuinely good.

The gap shows up precisely at the point AI is supposed to pay for itself: deflection at volume. Every platform in this category makes one of these moves - meter the AI, meter the document, meter the seat, meter the resolution. The "perfect" model for a buying team is the one that doesn't meter success at all.

What to look for instead

• A flat monthly price where AI is included, not a credit allowance that resets
• No per-conversation or per-resolution AI metering
• No per-seat fees as your team grows
• Multi-LLM routing, so routine questions go to a cheap model and hard ones to a flagship
• A free tier where you can run real traffic before you pay

How Chatonn prices AI support instead

Chatonn uses flat monthly plans with AI included - no credit meters, no per-seat fees, no per-resolution charges. Pro is $49/month and covers AI-managed conversations with no API keys needed, auto cost-routing to the cheapest capable model across OpenAI, Gemini, Claude, Groq, or self-hosted Ollama, unlimited team members, and 50 knowledge sources. Business is $149/month and adds white-label branding, 3 workspaces, unlimited knowledge sources, advanced analytics, and API access.

The point of the model is that success doesn't bill you. Better deflection lowers your support cost instead of draining an AI allowance - and your budget stays a single predictable line item.

Want to test it before you commit? Bring your own keys on the Free plan at chatonn.com and get 500 conversations a month at no cost. Compare plans on the pricing page - and see the full Crisp comparison on our Crisp alternative page.