Decagon

What it does

Done-for-you service

NoSoftware the buyer's own CX team builds and iterates on; the vendor positions itself against professional-services delivery.Instead of complex SDKs or costly professional services, you can define agent behavior in natural language, the same way you train human agents with…decagon.ai/product/aop ↗

How much it does unattended

Runs autonomously

YesAgents resolve customer conversations end to end and escalate to a human only when needed.At Decagon, our AI agents resolve customer service tickets autonomouslydecagon.ai/blog/pricing-ai-agents ↗

Multi-agent

PartlyA separate supervisor model reviews responses and Duet iterates on the agent, but the vendor describes a network of specialised models rather than agents coordinating on one task.A built-in supervisor detects hallucinations before responses are sent, automatically revising responses that stray from factual grounding.decagon.ai/security ↗

Agent permissions

PartlyGuardrails gate sensitive actions such as identity verification and refunds, and Duet's agent changes are staged for human review before release.Ensure critical operations like identity verification and refunds are executed securely, reliably, and on-brand with strict controls.decagon.ai/product/overview ↗

Which models it runs on

Claude

PartlyAnthropic is named as one of the AI providers behind the platform, but the vendor selects models and says most traffic runs on its own.Decagon enforces zero-day retention with all AI providers like OpenAI and Anthropic, ensuring no conversation data is stored or used for training.decagon.ai/security ↗

GPT

PartlyOpenAI is named as one of the AI providers behind the platform; model choice sits with the vendor.Decagon enforces zero-day retention with all AI providers like OpenAI and Anthropic, ensuring no conversation data is stored or used for training.decagon.ai/security ↗

Model choice

NoThe vendor picks the models: mostly its own post-trained models, orchestrated as a network, with no customer-facing model selection described.Today, over 80% of all model traffic at Decagon runs on models we trained ourselves. They outperform the best foundation models on our real-world use…decagon.ai/blog/introducing-decagon-labs ↗

Not established: Open models

Where you use it

In your pipeline

YesAgent versions can be tested inside existing CI/CD practices before release.Integrate with CI/CD practices to test every agent version and ensure every change meets the quality bar without regressions.decagon.ai/product/testing-qa ↗

In a browser

YesA hosted platform users sign into, with role-based access, where teams build, test and monitor agents.Platform access is defined and enforced by role, ensuring every user automatically gets the right permissions for their job.decagon.ai/security ↗

Not established: On the command line

Whose machine it runs on

Self-hosted

NoVendor-hosted SaaS on Google Cloud with logical tenant isolation; no customer-infrastructure deployment is offered, only EU-only residency on request.Decagon's primary production infrastructure is hosted on Google Cloud Platformdecagon.ai/legal/dpa-security-annex ↗

Open source

NoThe agent engine is described as proprietary; no source is offered.Every Decagon agent is built on a proprietary engine that understands your business and becomes more intelligent with every interaction.decagon.ai/product/overview ↗

What it costs to run

How it meters

YesMetered per conversation or, alternatively, per resolved conversation with no charge for escalations; not per seat.Per-conversation pricing: A fixed rate for every incoming conversation, with flexible pricing for higher volumes.decagon.ai/blog/pricing-ai-agents ↗

Free tier

NoTwo metered commercial models are the only ways in and every call to action is a demo request; no free or self-serve tier is offered.We've successfully deployed AI agents across countless customers and support two flexible pricing options:decagon.ai/blog/pricing-ai-agents ↗

API access

YesSelf-serve APIs let the agent fetch data and trigger actions, and customer data can be retrieved or deleted through publicly accessible APIs.Let your AI agent retrieve data and trigger actions through self-serve APIs and custom tool integrations.decagon.ai/product/integrations ↗

MCP server

PartlyMCP is supported for the agent to reach out to other systems; the vendor does not state that Decagon exposes an MCP server another agent could query.Connect to any data system or application via an open standard for real-time data and actions.decagon.ai/product/integrations ↗

Not established: Bring your own key

Buying it for a team

A company can buy it

YesSold to enterprises through sales, with DPAs executed with customers and subprocessors.Decagon executes Data Processing Agreements with customers and subprocessors that document data protection obligations and accountability…decagon.ai/legal/dpa-security-annex ↗

Seat model

NoExplicitly not seat-based: the vendor argues seats misprice agents and meters conversations or resolutions instead.SaaS pricing typically relies on the number of seats—the more seats, the higher the price. AI agents, however, are not "tools" for humans.decagon.ai/blog/pricing-ai-agents ↗

Admin controls

YesRole-based access control plus configurable guardrails and guidelines constrain what the agent and its operators may do.Set rules for brand voice, escalations, and hallucinations across across chat, voice, and email.decagon.ai/product/aop ↗

Audit log

YesTamper-protected logs of user activity, privileged administrator actions and security events, retained at least twelve months.Comprehensive, tamper-protected logs capture key events, from logins to unusual activity, accessible only to senior engineering leadership.decagon.ai/security ↗

Single sign-on

YesSSO with identity providers including Okta and Microsoft Entra.Decagon integrates with identity providers like Okta and Microsoft Entra, enabling secure access across systems without additional passwords.decagon.ai/security ↗

What happens to your code

Opt out of training

YesNo opt-out needed: model providers are under zero-data-retention agreements and customer data is never used for training or fine-tuning.Customer data is never used to train or fine-tune any AI models.decagon.ai/legal/dpa-security-annex ↗

Data residency

PartlyDefault processing is US-hosted on Google Cloud; EU-only residency inside EEA infrastructure is available on request.Decagon offers EU-only data residency, in which Customer Personal Data is processed and stored within EEA-located infrastructure.decagon.ai/legal/dpa-security-annex ↗

Getting out

YesCustomer data can be retrieved or deleted via public APIs at any time, and is deleted irrecoverably after termination; no credit balances to expire.Customer Personal Data submitted to the Services by Customer may be deleted or retrieved by the Customer using Decagon's publicly accessible APIs at…decagon.ai/legal/dpa-security-annex ↗

Certifications

YesAnnual SOC 2 Type II audits and ISO 27001 certification, with GDPR and CCPA/CPRA rights handling and annual third-party penetration testing.Decagon undergoes annual SOC 2 Type II audits covering the Security and Confidentiality Trust Service Criteria, and maintains ISO 27001 certification.decagon.ai/legal/dpa-security-annex ↗