Back to vendors
S

Scaled Cognition

Visit site
Entry priceContact for pricingFull pricing detail

CX model lab whose APT model powers policy-enforced agents that act on live systems, with a platform to author, simulate, score, deploy (hosted, VPC or on-prem) and monitor them.

Scaled Cognition is an AI model lab focused on customer experience. Its model, APT, is built for agents that take real actions in customer conversations, such as checking balances, changing orders or processing transfers, and is designed to follow business policy through architectural constraints instead of prompts. It validates every parameter against the customer's system schemas before calling a function, retrieves facts from live systems instead of generating them, and confirms that an action succeeded before telling the customer it is done.

Around the model sits a platform for building and running CX agents. Business teams can author in plain language, product teams in a browser-based Agent Builder, and engineers in a Python SDK, with every route compiled into deterministic specifications, and AgentTwin can turn transcripts of a company's best human agents into a working agent.

Before launch, GenAPI simulates backend systems, generative virtual customers stress-test agents across routine, edge-case and adversarial scenarios, and scorecards measure correctness, policy adherence and reliability across repeated runs. In production, Agent Defender blocks jailbreaks and policy violations, every action, blocked decision and policy check is logged, and monitoring tracks drift and unexpected behavior.

Agents can run hosted, in a customer's VPC or on-premises, with stateless processing and zero data retention, and the company is SOC 2 Type 2 compliant. Genesys uses Scaled Cognition's models to power virtual agents on Genesys Cloud. Customers can build independently or work directly with the company's engineers from proof of concept to production.

Vendor details

Canonical URL

https://scaledcognition.com

Category

Enterprise operations agent

Funding status

Private company.

Company status

independent

Use cases & customers

Deployment options

SaaScloudon-prem

In practice

A bank wants an agent that can move money for callers without ever acting on a wrong account. Scaled Cognition's APT checks every parameter against the bank's API schemas and confirms the transfer succeeded before telling the customer, running inside the bank's own VPC.

A contact center wants to launch an agent without testing on real customers. The team builds it in the Agent Builder, runs it against simulated backend systems and thousands of generated customers, and ships only when the scorecards show it follows policy.

An airline has transcripts of its best service agents handling rebookings. AgentTwin turns those transcripts into a working agent, and every action it takes in production is logged with what was allowed, blocked and why.

Agentic Index coverage score

8.0 / 14 capabilities · 57%

Integrations & Tool Calling Full

Typed actions against the customer's own API schemas connect agents to the customer's real systems. Teams "connect real API schemas so agents execute with verified contracts", and the Agent Builder lets them "test against simulated APIs... then deploy to real systems".

APT validates "every parameter against your system schemas before executing any function" and "confirms successful system execution" before telling a customer an action is complete, with actions on account balances, order statuses and transfers. Typed APIs specify exactly what each agent can read, write or modify. Custom tools are the customer's own API schemas, not only prebuilt connectors.

There are no named native connectors and no credential rotation or revocation. Genesys Cloud offers the model to its own customers through a partnership.

Sourcescaledcognition.com/product/explore-platform, /product/meet-apt-1; readread 2026-09-15

Workflow Orchestration Full

Agent behavior compiles into deterministic specifications that mix fixed conditions with agent steps. Teams author behavior as structured instructions in natural language, in a low code builder or in the Python SDK, "all compiled into deterministic specifications instead of prompts".

Deterministic conditions and function conditions "specify... preconditions that must be satisfied before APT is permitted to predict an action", so each action waits on its conditions and the flow branches at each step. Business, product and engineering teams all author on one deterministic core.

Changes go through what Scaled Cognition calls "controlled evolution", with "versioned deployments, evaluation gates, and validation", and execution is auditable and debuggable.

There are no explicit retry or fallback paths.

Sourcescaledcognition.com/product/explore-platform, /product/meet-apt-1; readread 2026-09-15

Knowledge Grounding & RAG Partial

Agents ground their answers in live data read from the customer's systems at run time, and there is no maintained index. APT promises "Zero guessing". It "retrieves facts, doesn't generate them" and grounds responses in real time information from the customer's systems, such as account balances, order statuses and policy details. Agents retrieve verified data, and the SDK enforces data provenance. Knowledge is read at run time, not trained in.

Each conversation reads what it needs through typed APIs at that moment. With "stateless processing, no storage, no training, no retention", nothing builds a lasting index over the customer's content. No sources are indexed or refreshed, and answers show the customer no citations.

Sourcescaledcognition.com/product/meet-apt-1, /product/explore-platform; readread 2026-09-15

Human Oversight & Guardrails Partial

The architecture enforces policy instead of a person signing off. Scaled Cognition promises "strict policy adherence", with "business rules enforced through architectural constraints, not prompts", so agents "physically cannot execute unauthorized actions". Function conditions set preconditions before any action. Agent Defender blocks jailbreaks, malicious prompts and policy violations in real time, before actions execute. APT "confirms execution, not intent".

Customers cannot choose which actions need a person's approval, who approves, or what happens when nobody does. There is no human approval step, approval routing or pause control, because enforced preconditions take the place of approval. Generative virtual customers can test escalations, and Scaled Cognition's own team verifies agents before go live as an engagement service.

Sourcescaledcognition.com/product/meet-apt-1, /product/explore-platform; readread 2026-09-15

Security, Identity & Governance Full

The platform is "SOC 2 Type 2 compliant", and its "controls are designed into the platform environment". Scaled Cognition runs a Vanta trust center at trust.scaledcognition.com. Under deterministic access control, typed APIs specify exactly what each agent can read, write or modify, and function conditions gate which actions are permitted. Platform tools process interactions ephemerally, with "no storage, no training, no retention".

People using the builder get no SSO, SAML, SCIM or user roles. The SOC 2 report's auditor and date are not public.

Sourcescaledcognition.com/product/go-live, /product/explore-platform, trust.scaledcognition.com (metadata only); readread 2026-09-15

Observability & Auditability Full

Every action, blocked decision and policy check is logged, and live agent behavior is monitored. The log provides "complete traceability for compliance, debugging, and oversight", and "every interaction within the platform is logged and auditable". Every agent action is traceable, "with clear visibility into what was allowed, blocked, and why", including the steps taken and the tools called. Behavior monitoring in live customer conversations catches "issues that aggregate metrics miss", and Agent Monitor tracks "drift, policy violations, and unexpected actions".

There is no SIEM export and no retention by plan. Platform tools run statelessly with zero data retention, and Scaled Cognition does not say how long the audit logs themselves persist. Trend & Topic Analysis reads conversation content for business strategy, not agent behavior.

Sourcescaledcognition.com/product/explore-platform, /product/go-live; readread 2026-09-15

Memory & State Persistence Not documented

No state carries between conversations, by design. Agents run "with stateless execution, zero data retention", and the platform promises "stateless processing, no storage, no training, no retention". Platform tools process interactions ephemerally. There is no memory beyond the conversation in progress. AgentTwin learns from historical transcripts, but only to build the agent's specification.

Sourcescaledcognition.com/product/explore-platform; readread 2026-09-15

Deployment & Data Residency Full

Agents run hosted, in the customer's own VPC or on premises, all "with stateless execution, zero data retention". Customers can "deploy on your infrastructure (hosted, VPC, or on-prem) to maintain complete data sovereignty", and in a VPC or on premises sensitive data stays inside the customer's environment.

Scaled Cognition does not say which features each mode supports or who patches in each mode, and customers cannot choose a hosted region.

Sourcescaledcognition.com/product/explore-platform, /product/go-live; readread 2026-09-15

Prebuilt Agents, Templates & Packs Not documented

There are no prebuilt agents, templates or catalog. AgentTwin "learns from what your best agents already do, and ships it as a production-ready agent, straight from your transcripts". It builds a custom agent from the customer's own data, and nothing is prebuilt or picked from a list. The Vibe Builder, Studio and SDK are authoring tools. Scaled Cognition targets telecommunication, travel and hospitality, banking and finance, retail, BPO and CX service providers, and healthcare, with no packaged agent for any of them.

Sourcescaledcognition.com/product/explore-platform, sitemap; readread 2026-09-15

Triggers & Channel Coverage Partial

Live customer conversations reach the agents through Genesys Cloud, a third party's contact center platform, not through channels of Scaled Cognition's own. Genesys "has selected Scaled Cognition's specialized customer experience large action models to power virtual agents on the Genesys Cloud platform", and the inbound queue and channels in that deployment belong to Genesys Cloud.

Contact center deployments bring inbound traffic, but Scaled Cognition's own product has no channel (voice, chat, email, messaging), inbound queue, webhook, event or schedule that starts an agent.

Sourcescaledcognition.com/blog/announcing-our-partnership-with-genesys (16 October 2025), /product/explore-platform; readread 2026-09-15

Model Flexibility & Routing Not documented

Every agent runs on APT, Scaled Cognition's own model, with no choice of provider. Scaled Cognition is a model lab. It calls APT "the only CX model built on super-reliable intelligence" and compares it with general LLMs, and every authoring route is "powered by the same deterministic core: APT". There is no provider list, no routing policy and no way to bring your own model or key. Genesys Cloud offers APT to its own customers, but that adds no model choice inside Scaled Cognition's platform.

Sourcescaledcognition.com/product/meet-apt-1, /product/explore-platform; readread 2026-09-15

APIs, SDKs & MCP Extensibility Partial

The platform has a Python SDK, the APT SDK, but no published documentation for it. The SDK gives technical teams "engineering-grade control" to script, extend and integrate agents in Python, enforce data provenance and test behavior. It sits alongside the Vibe Builder and Studio.

The site has no developer page, API reference or SDK page, and the API host at api.scaledcognition.com sits behind an access gate. There is no MCP surface.

Sourcescaledcognition.com/product/explore-platform, /product/meet-apt-1, sitemap, api.scaledcognition.com probe; readread 2026-09-15

Testing, Debugging & Optimization Full

Agents are tested against simulated systems and customers, and the results are readable and comparable. GenAPI emulates the customer's infrastructure "without connecting to live systems", generative virtual customers simulate realistic and adversarial behavior, and a test case generator builds scenarios from standard workflows to edge cases.

Agent scorecards measure correctness, policy adherence and reliability across repeated runs, and track them over time. Evaluation runs agents against thousands of simulated scenarios "to catch performance regressions before deployment". Changes go through "controlled evolution", with versioned deployments, evaluation gates and validation, and Agent Monitor keeps checking agents after deployment.

Scaled Cognition's team also offers expert led evaluations as an engagement service, separate from the product.

Sourcescaledcognition.com/product/explore-platform, /product/meet-apt-1; readread 2026-09-15

Browser & Computer Use Not documented

Agents work through typed APIs, so there is no browser or desktop control. They act through "APIs & typed actions" with "verified contracts" and are tested against GenAPI simulations of backend APIs. APT validates every parameter against the customer's system schemas.

Sourcescaledcognition.com/product/explore-platform, /product/meet-apt-1; readread 2026-09-15

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

·Agent capabilityVerified

Scaled Cognition Voice, a text to speech model built for live customer calls, is generally available and streams audio through a standard API to the company's APT agents, third party agents, contact center platforms or legacy IVRs. The company states 70 ms latency and more than 25 languages, and voices can be cloned from a short recording. It runs self hosted in the customer's own cloud or data center at $1 per million characters, or hosted by Scaled Cognition at $9 per million characters.

Bears on: Deployment / data residency

View source
·Agent capabilityVerified

Scaled Cognition launched its Agentic Pretrained Transformer (APT) model and system. The platform is specifically designed and trained to power agentic applications and execute real world actions.

Bears on: Agent capability

View source
View all 2 Scaled Cognition updates →Tracked since Aug 2026 · Verified from public vendor sources

Pricing

Contact for pricing

Key ambiguities

No pricing page exists in the site map, and no first-party page states how cost scales, including whether it tracks automated interaction volume.

Agentic Index verified 2026-09-15

Alternatives to Scaled Cognition

The closest documented capability profiles to Scaled Cognition among enterprise operations agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.

  • HireVue7.5 / 14Adds documented Prebuilt Agents, Templates & Packs
  • GoodTime9.0 / 14Adds documented Prebuilt Agents, Templates & Packs
  • IBM Sterling9.0 / 14Adds documented Prebuilt Agents, Templates & Packs
  • MakinaRocks9.0 / 14Adds documented Prebuilt Agents, Templates & Packs and Model Flexibility & Routing
  • MarvelX9.0 / 14Adds documented Prebuilt Agents, Templates & Packs
  • Reclaim AI6.5 / 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

Contact us

Found a vendor we missed? Have feedback on the index? We'd love to hear from you.