Not for SMBs or teams that want transparent self-serve pricing, a public API, or a fast no-engineering setup.

Changed since our last check

Agent permissions: now Yes. It was Partly when we checked on 19 October 2026.
API access: now Partly. It was Yes when we checked on 19 October 2026.
Audit log: now Yes. It was Partly when we checked on 19 October 2026.

The facts, with their sources

What it does

Done-for-you service: PartlyYou buy and operate software, but the vendor positions itself against services-heavy rivals while still bundling a forward-deployed team with every engagement to reach production faster.Most enterprises go live in weeks rather than months. A forward-deployed team are included with every engagement to accelerate your team’s path to…decagon.ai/vs/sierra

How much it does unattended

Runs autonomously: YesAgents handle customer conversations end to end and take actions in connected systems, escalating to humans only when needed; changes to the agent itself still go through human review.Decagon’s AI concierge remembers every customer, understands their needs, and takes action at the right moment.decagon.ai/proactive-agents
Multi-agent: PartlyDecagon runs a layered system of specialised models with a supervisor checking responses, but the vendor does not describe customer-orchestrated teams of agents coordinating on one task.a network of specialized models, each responsible for a distinct function—identifying the end of user speech, executing workflows, detecting…decagon.ai/blog/introducing-decagon-labs
Agent permissions: YesPermissions are bounded by guardrails at three points plus a bad-actor screen and Watchtower review, and sensitive operations like refunds and identity checks are explicitly gated.Was Partly before our last check.Before launch, simulations and regression testing validate behavior; during conversations, a supervisor model checks responses in real time and a…decagon.ai/vs/sierra

Which models it runs on

Claude: PartlyAnthropic is named as one of the AI providers Decagon sends data to under zero-retention terms, but the buyer does not choose Claude and most traffic runs on Decagon's own models.Decagon enforces zero-data 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 an AI provider on the security page and listed as a subprocessor, so GPT models are in the stack without being a buyer-selectable option.decagon.ai/security
Model choice: NoDecagon picks the models: over 80% of its model traffic runs on models it trained itself, and no page offers the buyer a model selector or bring-your-own-model path.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

Where you use it

In your pipeline: PartlyAgent tests can be wired into CI/CD to gate every agent version, but the vendor does not claim the agent acts on pull requests in a code repository.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: YesBuilding, testing, analytics and iteration all happen inside Decagon's hosted web product, which is where CX teams work day to day.Decagon puts everything (building, testing, analytics, iteration) in the product, so your team improves the agent as quickly as your business moves.decagon.ai/vs/sierra

Whose machine it runs on

Self-hosted: PartlyOne FAQ sentence says the platform deploys headlessly inside your existing infrastructure, but no page documents a self-hosted edition, and hosting subprocessors are Google Cloud and AWS.agent logic and tools are Git-backed with versioned workspaces for staging and production, tools are written in Python with MCP support, and the…decagon.ai/vs/sierra
Open source: NoThe agent engine is described as proprietary and no repository, licence or source distribution appears anywhere on the vendor's site.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: YesBilling is consumption-based on conversations or resolutions rather than seats, and the vendor says most enterprises pick per-conversation for predictable costs.Pricing model | Per-conversation or per-resolutiondecagon.ai/vs/sierra
Free tier: NoEvery call to action is a demo or sales contact, deployment runs as an engagement with a forward-deployed team, and no self-serve signup, free plan or trial is published anywhere on the site.decagon.ai/vs/sierra
API access: PartlyThe vendor advertises self-serve APIs and custom tool integrations so the agent can fetch data and trigger actions, but the API documentation portal is behind a login, so public programmatic access is unverifiable.Was Yes before our last check.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 in the outbound direction — Decagon's agent connects to your systems over the open standard — but no public MCP server is documented for another agent to query Decagon itself.MCP open connectivity: Connect to any data system or application via an open standard for real-time data and actions.decagon.ai/product/integrations

Buying it for a team

A company can buy it: YesThis is sold to organisations only, through sales engagements with named enterprise customers; there is no individual purchase path on the site.Our platform enables leading enterprises like Avis Budget Group, Chime, Oura Health, 1-800-FLOWERS.COM, and Hunter Douglas to deploy AI agentsdecagon.ai/about
Seat model: NoThe published pricing model is consumption-based per conversation or per resolution, so there is no seat price, seat floor or seat ceiling stated for the platform.Sierra prices per outcome (per resolution), while Decagon lets you choose per-conversation or per-resolution pricing.decagon.ai/vs/sierra
Admin controls: YesAdmins constrain both people and the agent: platform access is enforced by role, and guardrails plus guidelines govern escalations, refunds and identity verification.Platform access is defined and enforced by role, ensuring every user automatically gets the right permissions for their job.decagon.ai/security
Audit log: YesTamper-protected audit logs capture logins and unusual activity, though the vendor states access to them is restricted to senior engineering leadership rather than the customer's admins.Was Partly before our last check.Comprehensive, tamper-protected logs capture key events, from logins to unusual activity, accessible only to senior engineering leadership.decagon.ai/security
Single sign-on: YesSingle sign-on is offered against enterprise identity providers including Okta and Microsoft Entra; the vendor names the providers rather than the SAML or OIDC protocol.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: YesZero-data retention is enforced with all model providers by default, so conversation data is neither stored by them nor used to train their models; no opt-in is required.decagon.ai/security
Data residency: PartlyThe trust centre lists US and EU regions for its hosting and model subprocessors and the security page claims multi-region infrastructure, but no page names a guaranteed customer-selectable region.Google Cloud Platform • Cloud hosting provider US, EU Infrastructure Provider providing hosting services, storage and AI servicestrust.decagon.ai
Getting out: PartlyMemory is advertised with full data portability and a trust-centre control states customer data is deleted on departure, but no export mechanism or credit-expiry term is documented.Complement your systems of record with full data portability, while retaining granular control over what context is stored and how it’s used.decagon.ai/proactive-agents
Certifications: YesThe trust centre publishes SOC 2 Type II, ISO 27001:2022, GDPR, CCPA, HIPAA, PCI DSS and EU AI Act compliance, with the SOC 2 report and ISO certificate downloadable on request.trust.decagon.ai

Plans and pricing

Price and terms

Free plan or trial: NoNo

Checked 21 October 2026.

Works with

Claude
Partly
GPT
Partly

For AI agents

MCP server: Partly
API: Partly

AdYour company across The Agents IndexAdvertise here

Filed under

Support agentsAPIAutonomousEnterpriseNo-code

Sources and updates

Last changed 21 October 2026

Sources16

Coverage elsewhere7

The Agents Index in AI answers0 of 1

PerplexityDoes not cite The Agents IndexAsked Decagon alternatives on 13 October 2026Other sources: wotnot.io, g2.com, getmacha.com, itechguides.com, chatbase.co, ringly.io

Updates6

21 October 2026Free plan · Usage terms · What it can do
19 October 2026Free plan · Price · Usage terms · What it can do
18 October 2026Free plan · Price · Usage terms · What it can do
14 September 2026Free plan · Price · Usage terms · What it can do
2 September 2026Free plan · Usage terms · What it can do
31 August 2026Free plan · Usage terms · What it can do

Is this yours? Claim it free and edit it in your dashboard, or feature it in Support agents.