Sierra

What it does

Writes code

PartlyGhostwriter turns a prompt, an SOP or a transcript into a production agent journey with workflows and guardrails; what it authors is agent configuration, not application source code.Build or modify agents by describing how you want them to behave. Update workflows, systems integrations, guardrails, tone, and style with simple…sierra.ai/product/ghostwriter ↗

Debugs and fixes

PartlyGhostwriter diagnoses its own failed simulations and implements the fix; the failure it repairs is agent behaviour, never a defect in your software.When a simulation fails, Ghostwriter diagnoses the issue and implements the fix, so your agent keeps improving without manual debugging.sierra.ai/product/ghostwriter ↗

Writes tests

PartlyConversation and voice simulations are generated and run on every agent build, and Agent Studio keeps them as a regression suite; these test agent behaviour, not your code.Ghostwriter proactively runs tests with every build or update, so each change is validated before it reaches you.sierra.ai/product/ghostwriter ↗

Ships it

PartlyThe platform deploys what it builds, and one agent goes live across every channel, so work does not stop at a proposed change; it ships agents, not code.Build once and deploy across any channel—chat, phone, email, SMS, and messaging.sierra.ai/product/agent-sdk ↗

Done-for-you service

PartlyYou configure software in Agent Studio or the SDK, but Sierra staffs the engagement with its own agent development team and keeps optimising after launch.Work with an expert agent development team that has supported hundreds of deployments — and will keep you ahead as AI evolves. Only pay for the value…sierra.ai/product ↗

How much it does unattended

Runs autonomously

YesHorizon agents plan and execute toward an outcome across days or months without a person driving each step, and the Horizon terms warn that actions occur without human review of each one.Horizon acts on every signal and decides what's next — staying with the work until it's done, across days or months.sierra.ai/product/horizon ↗

Multi-agent

YesNamed agents pass work to each other: Explorer analyses conversations and hands a recommendation to Ghostwriter, which implements it.Send any recommendation directly to Ghostwriter, and it implements the improvement automatically, without any manual work on your part.sierra.ai/product/explorer ↗

Agent permissions

PartlyAutonomy is configured, not fixed: messaging is fully autonomous by default while you declare which steps need human sign-off, and unresolved cases hand off to your care team.Define what the agent can do on its own, and where human sign-off is required.sierra.ai/product/horizon ↗

Which models it runs on

Claude

PartlyAnthropic is a named subprocessor for large-language-model serving inside the platform, but Sierra routes each turn itself, so choosing Claude is not something a customer does.We are adding the following entities to our list of Subprocessors: Google, Anthropic, and an additional processing purpose for AWS. These three…trust.sierra.ai/updates ↗

GPT

PartlyOpenAI is listed as an AI-services subprocessor with US and EU processing, and Sierra publishes an agent channel inside ChatGPT; the customer does not select the model.OpenAI, L.L.C. AI services Global (localization available in United States, European Union)trust.sierra.ai/subprocessors ↗

Open models

YesOpen-weight models are part of the mix Sierra runs alongside frontier and its own proprietary models; Sierra decides which serves each turn.Sierra is built on a constellation of LLMs—combining frontier, open-weight, and proprietary models—so your agent can execute seamlessly and reliably.sierra.ai/product/trust-and-reliability ↗

Model choice

NoSierra picks the model and switches providers on its own for performance and failover; no page describes a customer-facing model choice.Sierra automatically switches between LLM providers to optimize your agent’s performance and maintain service continuity in case one provider…sierra.ai/product/trust-and-reliability ↗

Where you use it

In your editor

PartlyAgent SDK journeys are written in your own editor and software lifecycle, but Sierra ships no editor extension and the agent itself runs in customer channels.Develop agents in your existing programming environment and software development lifecycle.sierra.ai/product/agent-sdk ↗

In your pipeline

PartlyAn agent release can be pushed from your CI/CD pipeline, but the agent does not act on pull requests or run build jobs.Make your agent available on ChatGPT with the push of a button or via CI/CD.sierra.ai/product/channels ↗

In a browser

YesThe agent runs as hosted chat on your website and mobile app, alongside WhatsApp, Apple Business Chat and SMS.Power chat conversations across your website, mobile app, WhatsApp, Apple Business Chat, and SMS.sierra.ai/product/channels ↗

Not established: On the command line

Whose machine it runs on

Self-hosted

NoSierra hosts the platform on its own AWS infrastructure; no customer-hosted, on-premise or bring-your-own-cloud deployment appears anywhere on the public site.Amazon Web Services, Inc. Data storage, cloud hosting, and AI servicestrust.sierra.ai/subprocessors ↗

Open source

NoProprietary platform: the terms reserve every right in the Services to Sierra and separately prohibit decompiling or reverse-engineering the software.Sierra and its licensors exclusively own all right, title and interest in and to the Services, including all associated intellectual property rights.sierra.ai/terms-and-conditions ↗

What it costs to run

How it meters

YesBilled per delivered outcome such as a resolved conversation, a saved cancellation or an upsell, with escalated cases usually free, and consumption pricing blended in where an outcome cannot be attributed.With outcome-based pricing, Sierra gets paid only when we complete a task for you.sierra.ai/blog/outcome-based-pricing-fo… ↗

Free tier

NoNo free tier, trial or self-serve signup: sign-in requires a company deployment that already exists, and every call to action is a contact form.Enter your company's Sierra URLsierra.ai/login ↗

API access

YesA headless API is one of the published deployment surfaces; the reference is behind the docs sign-in, so endpoints, auth and limits are not public.Because ADP is powered by Agent OS, you can build once and deploy everywhere you are — chat, phone, SMS, email, contact center conversations, or via…sierra.ai/blog/agent-data-platform ↗

MCP server

YesAgents are exposed to ChatGPT Apps over MCP, and you choose which journeys, data and capabilities each channel sees.Sierra agents are natively compatible with ChatGPT Apps via Model Context Protocol (MCP).sierra.ai/product/channels ↗

Bring your own key

NoSierra supplies and routes the inference itself through its own model subprocessors; no bring-your-own-key, own-model or own-cloud option is published.Sierra will power your AI agent with the best available LLMs in the industry, so you (and your customers) are always benefiting from the latest…sierra.ai/product/trust-and-reliability ↗

Buying it for a team

A company can buy it

YesEnterprise-only purchase, contracted per company under a Platform Services Agreement and Order Form; there is no individual plan.40% of the Fortune 50 partner with Sierra.sierra.ai/product ↗

Seat model

NoNo seat licences at all, so there is no floor or ceiling to negotiate; Sierra argues seat pricing conflicts with automating the work.With no reliance on seat-based pricing, we have no conflicting incentives—our success is fully aligned with your outcomes.sierra.ai/blog/outcome-based-pricing-fo… ↗

Admin controls

YesGoals, guardrails and topic and keyword filters bound what the agent may say or do, and Horizon lets you mark the steps that need human sign-off.Sierra has built-in filters and monitors for topics and keywords that are off-limits based on your company policies.sierra.ai/product/trust-and-reliability ↗

Audit log

PartlyEvery agent decision, tool call and knowledge lookup is traced and reviewable, but Sierra publishes nothing about logging administrator or configuration changes.Understand the reasoning behind every agent action or answer—view knowledge sources, systems accessed, and more.sierra.ai/product/insights ↗

Not established: Single sign-on

What happens to your code

Opt out of training

YesSierra states customer data is never used to train models and is only ever used as you instruct, so there is no opt-out to negotiate.Your data is never used to train models, and we use industry-standard best practices to ensure its security.sierra.ai/product/meet-your-agent ↗

Data residency

PartlyHosting can be localised to the United States, European Union, Southeast Asia, Japan or Australia, but model-serving subprocessors localise to the US and EU only and no product page explains how a region is chosen.Global (localization available in United States, European Union, Southeast Asia, Japan, Australia)trust.sierra.ai/subprocessors ↗

Getting out

YesConversation memory is yours to export to your warehouse or any external system, retained on terms you set; there are no credits that could expire.Own your memory layer—permissioned, retained on your terms, and exportable to your warehouse or any external system.sierra.ai/product/context-engine ↗

Certifications

YesSOC 2 Type II, ISO 27001, ISO 42001, HIPAA, PCI DSS 4.0.1 Level 1 Service Provider, GDPR, CCPA, CSA STAR, EU AI Act and FedRAMP High; audit reports are released from the Trust Center on request.Sierra is committed to maintaining the highest compliance standards for our customers, including SOC 2, HIPAA, GDPR, PCI, FedRAMP High, CCPA, CSA…sierra.ai/product/trust-and-reliability ↗