Decagon review
★ 4.1/5AI support agent
Enterprise AI customer service platform branded "the AI concierge for every customer." It deploys AI agents across chat, email, and voice, configured with natural-language Agent Operating Procedures (AOPs) instead of code. Used by large brands like Notion, Duolingo, Eventbrite, and Rippling to deflect and fully resolve high support volume.
Affiliate link · how we make money
Best for enterprise customer support. Decagon starts Custom (~$50,000/yr platform fee + usage) on a usage-based, custom · per-conversation or per-resolution + platform fee model. Decagon does not publish prices, but using widely reported figures: a mid-market deployment handling 30,000 conversations/month on the per-conversation model at ~$0.99 each is ~$29,700/mo in usage (~$356,400/yr), plus the ~$50,000/yr platform fee, totaling roughly $400,000/yr. This lines up with third-party Vendr data showing a median annual contract of ~$432,750 (range ~$105,000 to ~$923,000). On the per-resolution model you pay a higher rate only for tickets the AI fully solves, so the bill tracks resolution rate rather than raw volume. Either way, voice and complex integrations push the quote higher, and the platform fee applies on top no matter which model you pick.
Spec
The gotcha
There is no published price and no entry tier under six figures. On top of usage you pay a recurring ~$50k+ annual platform fee plus onboarding/professional services, and a hard practical floor means contracts below ~$50k are not even a fit for their sales process. On per-resolution billing, watch how "resolution" is defined in the contract, since a loose definition can inflate the bill and make savings hard to predict.
Decode the real cost →Best for
- You want an enterprise-grade AI agent to fully resolve customer support across chat, email, and voice, you have high ticket volume and a six-figure budget, and you need deep integrations into Salesforce/Zendesk/Intercom plus a sales-led, white-glove rollout rather than a self-serve tool.
source: decagon.ai/pricing →
Decagon pricing tiers
Prices move. Cells flagged verify link to the live vendor page.
| tier | monthly | annual | included | unit |
|---|---|---|---|---|
| Per-conversation | Customverify | Custom | Fixed rate per incoming conversation handled by the AI; flexible/discounted pricing at higher volumes. Reported ~$0.99 per conversation. Decagon's most popular model. | per conversation |
| Per-resolution | Customverify | Custom | Higher fixed rate per fully resolved conversation, with no charge for escalations to humans. Rate not publicly disclosed. | per resolution |
| Platform fee | Customverify | $50,000 | Recurring annual platform fee charged regardless of which usage model is chosen; reported at roughly $50,000+/yr. Plus professional services/onboarding set during the sales process. | per year |
// pools and per-unit rates are volatile · cells flagged verify link to the live vendor page
Compare Decagon with…
Questions
How much does Decagon cost?
Decagon starts Custom (from ~$50,000/yr platform fee + usage). Decagon does not publish prices, but using widely reported figures: a mid-market deployment handling 30,000 conversations/month on the per-conversation model at ~$0.99 each is ~$29,700/mo in usage (~$356,400/yr), plus the ~$50,000/yr platform fee, totaling roughly $400,000/yr. This lines up with third-party Vendr data showing a median annual contract of ~$432,750 (range ~$105,000 to ~$923,000). On the per-resolution model you pay a higher rate only for tickets the AI fully solves, so the bill tracks resolution rate rather than raw volume. Either way, voice and complex integrations push the quote higher, and the platform fee applies on top no matter which model you pick.
Can you self-host Decagon?
No, Decagon is cloud-only.
Does Decagon support MCP?
Decagon supports MCP as both client and server, so it works with assistants like Claude and ChatGPT.
What's the catch with Decagon's pricing?
There is no published price and no entry tier under six figures. On top of usage you pay a recurring ~$50k+ annual platform fee plus onboarding/professional services, and a hard practical floor means contracts below ~$50k are not even a fit for their sales process. On per-resolution billing, watch how "resolution" is defined in the contract, since a loose definition can inflate the bill and make savings hard to predict.