Dante AI Review 2026

How we evaluated this

We evaluated Dante as a website knowledge agent with a flat subscription and a credit meter, not as a contact center. Scoring weights pricing honesty (flat fee versus credits and add-ons), model flexibility, handover packaging, knowledge freshness by plan, and implementation risk.

We read the official homepage and pricing page on August 19, 2026, including Free through Pro, yearly two-months-free copy, add-ons, and the credit FAQ. Vendor docs describe WhatsApp handover into the same inbox as the web widget. We did not run a production pilot or audit the 99.9% SLA.

Pricing calculation to run

Flat fee beats per-resolution only if you respect the credit ladder

Dante's comparison is fair as far as it goes. If a support org resolves 2,000 AI-handled tickets a month at $0.99 each, that is $1,980 in resolution fees before seats. Dante Pro is $400 per month plus credits you already paid for inside the plan. The catch is that "already included" is a credit pool, not an unlimited reply grant, and unused credits do not roll over.

The pricing FAQ on August 19, 2026 publishes a per-reply ladder:

• 1 credit: GPT-4o mini, GPT-4.1 nano, GPT-5.4 nano.
• 2 credits: Gemini Flash, GPT-4.1 mini.
• 5 credits: Gemini 3 Flash, GPT-5.4 mini, Claude Haiku 4.5, GPT-5.6 Luna.
• 10 credits: most frontier models (GPT-5, GPT-4.1, GPT-4o, Gemini 3.1 Pro, GPT-5.4, GPT-5.6 Terra).
• 15 credits: Claude Sonnet 4.6.
• 25 credits: Claude Opus 4.6 / 4.7, GPT-5.5, GPT-5.6 Sol.

Starter includes 2,000 credits. That is 2,000 nano replies, or 200 ten-credit frontier replies, or 80 Opus-class replies. Advanced's 8,000 credits is 800 frontier replies. Pro's 30,000 is 3,000 frontier replies or 1,200 at 25 credits. If you default every conversation to Opus because a demo looked fluent, Starter dies in a week. If you ship nano and only escalate hard questions to Sonnet, Starter can last a modest site.

Add-ons change the "flat" story. Extra credits are $20 per 1,000, which is cheaper than Chatbase's $40 per 1,000 message credits, but Dante credits expire at month end. Extra agents are $20 per month. Branding removal is $50 per month on Starter, Advanced, and Pro (Pro already includes white-label in the plan bullets, so confirm you are not double-buying). Human handover is $50 per month per 500 conversations on Starter and Advanced, and bundled in Pro. A team that stays on Advanced at $120 to avoid Pro, then buys handover ($50) and branding ($50), is at $220 before extra agents or top-ups. Pro at $400 may be the cleaner package once those two add-ons are real requirements.

Bring-your-own OpenAI key is the other lever: LLM calls ride your OpenAI contract and skip Dante credits for those calls. Useful if procurement already has volume pricing. You still pay the Dante plan for the product around the model.

Who Dante AI fits best

Channel coverage to verify

The primary surface is a website embed you can place on sites you control. Starter lists "all integrations and deployment options" without naming them on the pricing tile. Vendor docs describe WhatsApp handover into the same inbox. Voice appears in Dante's own how-to guides as an input or agent mode to enable after the chatbot exists. Treat those as items to demo, not as a confirmed omnichannel inbox on every plan.

• Website widget: core path. Test mobile, custom launcher (Starter), custom CSS (Advanced), and white-label (Pro or $50 add-on).
• Custom domain: 1 on Starter.
• WhatsApp: documented handover path. Confirm BSP, Meta fees, and whether it is on your target plan.
• Voice: documented as a feature to turn on. Confirm channel versus in-widget microphone input. Those are different jobs.
• API: Pro. Required for authenticated app embeds that are not a public marketing site.

Workflow execution and handoff

The operating loop is: ingest site and files, pick a model, answer, capture leads (Advanced+), refresh knowledge on a schedule, escalate when the visitor asks for a person or the agent cannot answer. Humans take over in the inbox. Docs say WhatsApp threads land next to web chats when a number is connected.

That loop fails in two boring ways. First, Starter and Advanced buyers forget handover is a $50 / 500-conversation pack and go live with no live path. Second, Pro buyers turn on daily updates but never assign an owner for the knowledge-gap report Advanced already advertises ("top questions and knowledge gaps"). Auto-update will refresh a wrong page as faithfully as a right one.

• After-hours FAQ on a documentation-heavy site.
• Lead capture on Advanced, with a definition of a qualified lead before you turn it on.
• Agency white-label on Pro, or the $50 branding add-on on lower rungs, with a per-client agent at $20 each.

Pricing checked August 19, 2026

Dante AI pricing: Free to $400 per month, two months free yearly

Monthly tiles: $0, $40, $120, $400. Yearly: $400, $1,200, $4,000 (advertised as two months free; the FAQ also says "save up to 30% when billed yearly"). Add-ons stack monthly. Cancel-anytime language is on the page. Unused credits do not roll over. Top-ups are $20 per 1,000 credits on paid plans.

Source: dante-ai.com/pricing, August 19, 2026. Confirm taxes, add-on stacking, and whether Pro white-label means you can skip the $50 branding add-on.

Published add-ons

AI capability and model choice

Free is limited to light models (GPT-4o mini, GPT-4.1 nano, GPT-5.4 nano). Every paid plan opens the rest of the picker, including GPT-5.x families, Claude Haiku / Sonnet / Opus, and Gemini 3.x, per the FAQ. Heavier models cost more credits. Plan choice is volume, not access. You can test models side by side on your content and switch in one click, according to the same FAQ.

That flexibility is the product. It is also how teams blow Starter. Measure grounded accuracy and credits per resolved conversation on two models before you set a production default. If a 10-credit model is only slightly better than a 2-credit model on your corpus, the cheaper default funds more conversations and a cleaner Pro upgrade later.

Feature areas to verify

• Website and document training, character caps by plan.
• Model picker and on-screen credit use per reply.
• Analytics on paid plans; knowledge-gap signals on Advanced.
• Lead capture on Advanced.
• Custom launcher, CSS, domain, white-label by rung.
• Human handover inbox, including WhatsApp if you need it.
• API, SSO, audit logs, SLA language on Pro.
• Auto-update weekly (Advanced) or daily (Pro).
• Usage cap controls on Starter+.
• BYO OpenAI key.

Analytics and operating visibility

Starter adds an analytics dashboard. Advanced adds AI-identified top questions and knowledge gaps. Those views only matter if someone closes the gaps. Export or screenshot week-one baselines: credits per conversation, unknown rate, handover rate, and lead-capture count if you use it.

Because unused credits die at month end, analytics has a finance job too. If you consistently burn 1,500 of Starter's 2,000, stay. If you top up $20 x 3 every month, upgrade to Advanced. If you use 20,000 credits of Opus on Advanced, you are paying for the wrong model or the wrong plan. Dante shows credits per reply in the product, per the FAQ. Use that, not a gut feel.

Security, data handling, and compliance

Pro lists SSO, audit logs, and 99.9% uptime SLA language. The public pricing FAQ does not substitute for a SOC 2 report. Request DPA, subprocessors, retention, training-data policy, and region controls in writing. BYO OpenAI key changes where inference happens. Document that path with security if you use it.

• Define topics the agent must never answer: credentials, payment disputes, medical or legal advice.
• Map 1 to 5 seats to admin versus operator. Five Pro seats is still a small control plane.
• If you need HIPAA or a signed BAA, ask. It is not on the public Free-to-Pro tiles we captured.

What questions should you ask before buying Dante AI?

• How many credits did 50 of our real questions consume on nano, on a 10-credit model, and on Sonnet 4.6?
• Do unused credits ever roll over on annual billing, or is monthly forfeiture the rule in all cases?
• Is human handover included in our quote, or is it $50 per 500 conversations?
• If we are on Pro, do we still see the $50 branding add-on at checkout?
• Show WhatsApp takeover into the same inbox as the widget, on our test number.
• What does the 99.9% SLA actually credit if we miss it?
• Can we use our OpenAI key for production, and what then happens to Dante credit accounting?
• Who on our side owns weekly knowledge-gap cleanup?

What red flags should you watch for with Dante AI?

• A demo that only uses nano credits, then a production default on Opus.
• Going live on Starter with no handover pack and no human path.
• Annual commit before you have a week of credit-burn data.
• Buying extra agents for brands that share one messy corpus.
• Treating "no per-resolution fee" as "no variable cost." Top-ups and add-ons are variable.
• Assuming Free includes a durable published agent. Free is 100 credits per month after onboarding.

What are the best alternatives to Dante AI?

Compare on meter type (credits versus messages versus resolutions) and on where handover lives.

• SiteGPTChoose SiteGPT when you want message quotas, citations, and staged auto-refresh, and you can model the double-count of question plus reply.
• ChatbaseChoose Chatbase when you need procedures, optional voice, and a helpdesk view on Standard, and you can live with message credits plus add-ons.
• YourGPT AIChoose YourGPT when omnichannel actions are the job, not a flat website agent with a credit picker.
• TidioChoose Tidio when live chat plus a lighter bot is enough and you do not want a model-credit spreadsheet.
• GorgiasChoose Gorgias when Shopify order context is the system of record.

Workflow test

What Dante AI needs to prove in a real workflow

1. Train on the live site and the five documents that drive tickets.Score 50 real questions. Record grounded correct versus fluent wrong.
2. Measure credits per conversationon the model you will actually ship, then on one cheaper and one more expensive model. Unused credits will not wait until next month.
3. Turn on handover on the plan you will buy.If that means a $50 pack, buy it in the trial. A Pro-only demo of takeover is not evidence for a Starter rollout.
4. Compare 30-day costagainst SiteGPT message quotas and Chatbase credits plus add-ons, using the same sample set. Then pick monthly or yearly.

• SiteGPT review
• Chatbase review
• YourGPT AI review
• Tidio review
• Customer support AI agents buyer guide
• How to choose an AI agent platform
• Buyer scorecard

Claim and source ledger

Should you choose Dante AI?

Choose Dante when you want a website knowledge agent with a flat subscription, a written credit ladder, and no per-resolution line item, and when you will pick a default model using measured credits per correct answer. Free is for tests. Starter at $40 is for low-volume, light models. Advanced at $120 adds weekly refresh and lead capture. Pro at $400 is the first public bundle that looks like a serious operating plan: API, daily refresh, handover, white-label, SSO.

Skip Dante if you need a Shopify helpdesk, Intercom-native resolutions, or live takeover on the free tier. Skip it if nobody will watch the credit meter. Run the four-step test and a peak-month model, including $50 handover packs if you are not buying Pro, before you take the two-months-free yearly deal.