Decagon
AI agent platform purpose-built for customer support, with deep integration into support toolchains.
Decagon builds enterprise AI agents for customer support, positioning them as an AI concierge that resolves issues end to end, not a chatbot that only suggests articles. Its agents handle conversations across chat, email, voice, and SMS, drawing on a company's knowledge bases, help articles, product documentation, past conversations, and internal policies to answer accurately, and escalating to a human with full context when a request falls outside what they should handle.
The centerpiece of the platform is what Decagon calls Agent Operating Procedures, or AOPs. Instead of building rigid coded decision trees, customer-experience teams write support workflows in plain language, when to issue a refund, when to escalate, how to handle a particular complaint, and Decagon compiles those instructions into executable agent logic.
The effect is a division of labor: CX operators own the business logic and can change agent behavior without waiting on an engineering sprint, while engineers manage the underlying integrations and guardrails. Crucially, AOPs trigger real backend operations, so an agent can retrieve and update customer data, process a refund, cancel a subscription, or query a third-party system, not just talk.
Decagon unifies its channels under one intelligence layer with cross-channel memory, so an agent recognizes a returning customer and recalls earlier issues, and a conversation that starts in chat can continue by voice without the customer repeating themselves. Its voice agents handle real-time calls with interruption handling and brand-matched voices, and can also run outbound calls for things like renewals and follow-ups. Increasingly the company routes traffic through language models it has trained in-house on customer-support interactions, alongside leading foundation models.
For the teams operating these agents, Decagon adds tooling for reliability and oversight: monitoring and quality assurance, guardrails, a trace view that shows how a given procedure executed for a specific conversation, and tools that help debug and draft new procedures. It is an enterprise-focused, managed offering, deployed with the help of Decagon's own product managers and engineers instead of self-serve signup, and priced around resolved conversations, not per agent seat, aimed at large support operations in sectors like fintech, retail, travel, and consumer technology.
Vendor details
Canonical URL
https://decagon.ai
Category
Customer support agent
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
In practice
Your CX team wants to change how the agent handles refunds without filing an engineering ticket. Decagon's Agent Operating Procedures let them write the workflow in plain language, and the platform compiles it into executable agent logic.
A customer starts in chat, then calls back later. Decagon unifies chat, email, and voice with cross-channel memory, so the voice agent already knows the prior issue and the customer doesn't repeat themselves.
You need an agent that resolves issues, not just answers them. Decagon's agents take real actions like processing refunds, updating accounts, and querying your systems, and escalate to a human with full context when needed.
Agentic Index coverage score
10.0 / 14 capabilities · 71%
| Integrations & Tool Calling | Full |
|---|---|
|
Connectors span CRM, helpdesk and ticketing (Salesforce, Intercom, Zendesk), knowledge bases (Confluence, Contentful, Kustomer), and CPaaS and telephony (Amazon Connect, RingCentral, SIP trunking), plus custom endpoints and MCP connectivity for any other system. Agents both retrieve data and take action. Sourcedecagon.ai/product/integrationsread 2026-09-05 |
|
| Workflow Orchestration | Full |
|
Agent Operating Procedures turn natural language inbound and outbound workflows into executable agent logic that pulls live data and triggers actions across connected systems. Versions are tracked in Git, and rollout is staged across agent versions. Sourcedecagon.ai/product/aopread 2026-09-05 |
|
| Knowledge Grounding & RAG | Full |
|
Decagon maintains the agent's own knowledge base, filled by sync from customer systems including Confluence, Contentful and Kustomer. Suggestions keeps it current by finding gaps from conversations, ranking their impact on coverage and issuing monthly content updates. Traces name the knowledge articles the agent referenced. The knowledge base persists between conversations. Sourcedecagon.ai/product/suggestionsread 2026-09-05 |
|
| Human Oversight & Guardrails | Partial |
|
Guardrails can be configured for brand voice, escalation and hallucination. A supervisor model revises ungrounded responses before they are sent, bad actor detection escalates to a human, and Watchtower flags conversations for triage against criteria the customer sets. Oversight is automated gating and quality review after the fact, with no customer facing approval queue that holds an agent action until someone signs off. Sourcedecagon.ai/product/aopread 2026-09-05 |
|
| Security, Identity & Governance | Full |
|
Access controls include role based access control, SSO with Okta and Microsoft Entra, tamper protected audit logs and short lived scoped JWTs for agent access. Data is encrypted with AES-256 at rest and TLS 1.2+ in transit, DLP based redaction removes PII, and model providers are held to zero-day retention. Decagon shows SOC 2, ISO 27001, PCI DSS, HIPAA, CCPA and Data Privacy Framework marks and runs a live trust portal. SourceDecagon security page and trust.decagon.ai; decagon.ai/securityread 2026-09-05 |
|
| Observability & Auditability | Full |
|
Traces explain how and why the agent decided as it did at any point in a conversation, covering model calls, workflow triggers, the knowledge articles referenced, and runtime and latency figures. Watchtower reviews every conversation against criteria the customer defines and flags outliers for triage. Sourcedecagon.ai/product/overviewread 2026-09-05 |
|
| Memory & State Persistence | Full |
|
Memory is a named part of the agent engine. It holds customer context across sessions and channels alongside the customer's systems of record, with stated data portability and control, so a conversation started in chat continues by voice without repetition. Decagon does not publish how long memory is kept or how to delete it for one customer. Sourcedecagon.ai/product/overviewread 2026-09-05 |
|
| Deployment & Data Residency | Partial |
|
The platform runs on multi region infrastructure with autoscaling and automatic failover for resilience. There is no customer selectable region, VPC, single tenant or on premises option, so where data sits depends on Decagon's architecture, not the customer's choice. The product is delivered only as SaaS hosted by Decagon. Sourcedecagon.ai/securityread 2026-09-05 |
|
| Prebuilt Agents, Templates & Packs | Not documented |
|
Customers build agents by writing Agent Operating Procedures in natural language, or Duet generates them from the customer's own conversation transcripts, applying Decagon's deployment learnings behind the scenes. The seven industry pages are positioning and case studies, not solutions to adopt. There is no template gallery, named prebuilt agent set, blueprint library or industry pack. Sourcedecagon.ai/product/aopread 2026-09-05 |
|
| Triggers & Channel Coverage | Full |
|
Chat, voice and email run on one intelligence layer. Inbound calls route over SIP and CPaaS platforms, email comes in through Zendesk and Intercom, live chat escalates through Zendesk Sunshine and Salesforce, and outbound and proactive agents are supported as a core feature. Sourcedecagon.ai/product/overviewread 2026-09-05 |
|
| Model Flexibility & Routing | Partial |
|
The agent engine runs on a network of frontier and specialized models, with OpenAI and Anthropic named as providers held to zero-day retention, and model redundancy is a platform property. Routing happens inside Decagon, and customers and admins cannot select the model that powers the product. Sourcedecagon.ai/product/overviewread 2026-09-05 |
|
| APIs, SDKs & MCP Extensibility | Partial |
|
A REST API for the platform sits at api.decagon.ai, authenticated with a customer issued key passed as a bearer token, and rate limited. An API key can be generated from the Developer page of the Decagon dashboard. The documentation at docs.decagon.ai sits behind a customer login, nothing on the public site links to it, and the public Integrations page describes only connections into the customer's own systems. SourceDecagon developer documentation at docs.decagon.airead 2026-09-05 |
|
| Testing, Debugging & Optimization | Full |
|
Testing & QA and Experiments are customer facing products. Agents are validated before launch with end to end simulated conversations and unit tests. Live traffic is then routed across agent versions to A/B test changes and measure their effect on CSAT, deflection rate and other metrics at scale, with readable results the customer can compare. Sourcedecagon.ai/product/overviewread 2026-09-05 |
|
| Browser & Computer Use | Not documented |
|
The agent acts only through API and MCP connectors, CRM and ticketing integrations, CPaaS platforms and SIP telephony. It does not control a browser or virtual desktop or operate software that lacks a programmatic interface. Sourcedecagon.ai/product/integrationsread 2026-09-05 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Product updates
The Personal Agent Consent & Trust Protocol (PACT), which Decagon open sourced with Instinct, builds on the Agent2Agent protocol and OAuth 2.0 so a business can verify which personal agent is calling and what the customer has authorized it to do. The customer signs in with the business and approves named scopes such as orders:cancel, and replies carry signed receipts of the scopes used and actions taken.
Bears on: Human approval / guardrails
View sourceDecagon released Voice 3, a voice agent rebuilt so a fast conversational model listens and speaks while a stronger model handles reasoning, tools and guardrails in parallel. The agent can narrate progress and take follow up questions during long tasks, and it speaks through Chord, Decagon's own speech model for customer calls, switching automatically across more than 70 languages.
Bears on: Agent capability
View sourceDecagon introduced Browser Actions, a feature that allows its AI agents to access and complete tasks inside any web-based system. This capability enables the agent to navigate and interact with web interfaces directly.
Bears on: Browser/computer use
View sourcePricing
Contact sales
Usage, per conversation or per resolution. Per conversation is the default and bills every interaction, and per resolution charges a higher rate only for conversations the AI fully resolves.
What is public
Decagon publishes its pricing model but no price card and no figures. It sells an enterprise AI customer support platform whose autonomous agents resolve tickets end to end across chat, email, voice and SMS with cross channel memory. Agent Operating Procedures (AOPs) are natural language workflows, Voice 2.0 runs at sub second latency and adds outbound voice, Watchtower handles QA, and AI Actions work with Stripe, Shopify and Salesforce.
Billing mechanics
Usage based, with agents priced as workers rather than seats. There are two usage models. Per conversation is the default, and Decagon says "the majority of our customers gravitate towards per-conversation pricing"; it bills every interaction, resolved or not. Per resolution carries "a higher fixed rate for each fully resolved conversation, with no charge for escalations." Everything is custom quoted on annual contracts, with hands on onboarding included. Decagon has no native helpdesk and sits on Zendesk, Salesforce, Intercom or Kustomer.
Cost watchouts
The per conversation model charges for unresolved and escalated interactions too. What counts as a resolution can be unclear, and Decagon's own glossary says of resolution based pricing that "Gray areas can lead to billing disagreements." Decagon has no native helpdesk, so a Zendesk, Salesforce or similar license is a separate cost. Integration with custom ERPs carries professional services fees.
Variable cost rationale
Decagon meters conversations or resolutions. On the per conversation model, which most customers choose, every interaction is billed whether the AI resolves it or a human takes over, so the bill grows with total support volume. On the per resolution model the rate is higher but escalations cost nothing, so the bill grows with the number of conversations the AI fully resolves. No rate, cap or minimum is published. Volume discounts on larger commitments and the annual contract are what hold the unit cost down.
Additional watchouts
Decagon sells to enterprises only, with no public pricing and a sales led procurement. Per conversation billing pays for conversations the AI fails to resolve, and the definition of a resolution can lead to billing disputes.
Overage / add-ons
Usage is billed per conversation or per resolution, with volume discounts on larger commitments. Per conversation billing charges for interactions that escalate to a human too, while per resolution charges nothing for escalations.
Sales call required
Yes, required for paid access
Free / trial
None published
Lowest paid plan
Not published
Commercial notes
Founded in 2023 by Jesse Zhang and Ashwin Sreenivas. Decagon has raised about $481M, most recently a $250M Series D in January 2026 with Coatue and Index at a $4.5B valuation, triple its $1.5B valuation about six months earlier, followed by an employee tender at the same valuation in March 2026. It counts 100+ enterprise customers, among them Duolingo, Chime, Hertz, Affirm, Dropbox, Notion, Rippling, Oura and Classpass.
Key ambiguities
No figures are published. Decagon names its two usage models and says per resolution carries the higher fixed rate, but publishes no rate, platform fee, minimum or contract size.
Cancellation / refund
Annual contracts, sometimes multi year, with implementation included in the contract price. No self serve signup, and no cancellation or refund terms are published.
Support SLA / resale
Onboarding is hands on, with dedicated Agent PMs and Forward Deployed Engineers, and existing customers get Decagon University.
Missing data
Every figure: the per conversation and per resolution rates, any platform fee, minimums, contract sizes and volume discount tiers. The definition of a resolution is not published either, which leaves room for billing disputes.
Related vendors
- 5.Y — 5
- Ada — Agentic customer experience platform whose AI agents autonomously…
- Aisera — Enterprise agentic AI platform that orchestrates specialized agents…
- ASAPP — Generative AI for enterprise contact centers that plugs into…
- Assembled — AI customer support orchestration platform that unifies autonomous…
- BIK — Agentic AI CRM for eCommerce that consolidates marketing and…
Alternatives to Decagon
The closest documented capability profiles to Decagon among customer support agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- ASAPP11.5 / 14Adds documented Prebuilt Agents, Templates & Packs
- Ada11.0 / 14Adds documented Prebuilt Agents, Templates & PacksDecagon vs Ada →
- Lorikeet11.0 / 14Adds documented Prebuilt Agents, Templates & PacksDecagon vs Lorikeet →
- Nurix AI9.0 / 14Adds documented Prebuilt Agents, Templates & Packs
- Sinch12.0 / 14Adds documented Prebuilt Agents, Templates & Packs
- Delight.ai11.5 / 14Adds documented Prebuilt Agents, Templates & Packs
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded