StateSet Overview

StateSet is the iCommerce company. The StateSet iCommerce Engine is the execution layer for AI commerce — a safe, deterministic way for autonomous agents to operate real commerce across systems.

StateSet is the iCommerce company. iCommerce — intelligent commerce — is the agent-native execution layer for commerce, and the StateSet iCommerce Engine is that layer: it gives AI agents a safe, deterministic runtime for executing commerce operations across the systems a brand already runs.

The problem it exists for is specific. Critical commerce work still sits between systems — reconciling exceptions, applying policy, keeping operational state straight — and today that work is done by people. AI reasoning, agent protocols and buyer readiness have all advanced; the remaining bottleneck is trusted execution against real commerce systems. The Engine is built for that bottleneck: fewer human touches, faster resolution, lower operational cost, and autonomous execution you can audit.

Every part of the Engine exposes a REST API, and most parts also expose an MCP server and a skill. The intended operator is as likely to be Claude, or an agent you wrote, as a person in a console. If that is you, start at Getting started for AI agents.

What the Engine is made of

The Engine is one product with several parts. Read them as three layers plus the tooling that keeps agents accountable.

Commerce state — objects and transitions, not just text

Part	What it does	Reference
Commerce Engine	Orders, inventory, payments, returns, warranties, work orders, BOMs, purchase orders, invoices — the commerce objects and their allowed state transitions, with a stable set of invariant error codes an agent can branch on.	442 endpoints
Temporal Engine	Durable, long-running workflows — the ones that survive a retry, a restart, or a week-long wait for a supplier.	170 endpoints
NSR	Decision and proof-verification: what was decided, by whom, on what evidence — and a way for a third party to verify it.	259 endpoints

Systems of record — operate what you already have

StateSet is not another ERP. It sits above and across your systems of record and makes them safely operable by agents.

Part	What it does	Reference
Sync Server	Tenant-scoped order, inventory and product sync across the platforms a brand runs on — Shopify with NetSuite, a 3PL with the order book, inventory with every channel. Job-based, replayable, webhooks in and out.	86 endpoints
EDI	Trading-partner documents — the retailers and distributors that speak X12 and EDIFACT, translated to and from commerce state.	116 endpoints

Agent surfaces — resolve the interaction, execute the fix behind it

A support ticket is usually a symptom of an order, subscription or fulfillment problem. These are the parts that handle the interaction and close the operation behind it.

Part	What it does	Reference
ResponseCX	Agents with instructions, rules, knowledge, functions and evals, and an approvals queue for actions that move money.	ResponseCX API
Voice	The same agent model on the phone — Twilio and browser calls, supervisor monitor / whisper / barge, DNC and quiet-hours gates on outbound.	Voice API
Chat Widget	The embeddable front door: one required backend endpoint, a streamed event model, commerce outcomes captured in the analytics.	Chat Widget API
Computer Use	Execution in browsers and desktops, for the back-office tools that have no API.	Computer Use API
Sandbox	Isolated runtimes where an agent’s code runs — execute, files, tunnels, inference, long-running sessions with a cost cap.	Sandboxes

Keeping agents accountable

StateSet Agents is the evaluation and training tooling: it grades an agent’s production conversations against a reward, curates the ones that worked, retrains, and evaluates the result on prompts the model never saw — with lineage from the deployed model back to the conversations that produced it. It is how you show that an agent got better, not just that it changed.

How it is different

StateSet is not only an AI interaction layer or a workflow connector. It combines commerce semantics, deterministic execution, policies, integrations, auditability and outcome measurement.

Commerce-native

The Engine understands commerce objects and state transitions, not just text and API calls: orders, inventory, payments, returns, fulfillment, warehouse, finance.

Safe execution

Probabilistic reasoning bounded by deterministic validation, policies and approvals — invariants, auth, audit logs, simulation, replay, idempotency, fail-closed writes.

Operational breadth

The whole commerce lifecycle — CX, orders, subscriptions, returns, fulfillment, inventory, finance — instead of one interaction channel.

Interoperable

Works with the AI and protocol ecosystem you already use — MCP, embedded toolkits, multiple agent frameworks — rather than requiring a proprietary front door.

Embedded and deployable

Execution can run close to your infrastructure and data: customer-controlled deployment options, local and embedded runtimes.

Outcome accountable

Measured on the business work completed — successful autonomous commerce actions — not seats or messages.

Who it is for

If you are	The problem	What the Engine does
COO / Operations	Exceptions and cross-system handoffs scale with volume.	Removes the human glue between commerce systems and turns recurring operational work into governed autonomous execution — work off your team, not another dashboard.
Head of CX	Tickets are symptoms of order, subscription and fulfillment problems.	Resolves the interaction and executes the operational fix behind it, so closing the ticket does not create downstream work.
CFO	Ops and BPO cost rise with GMV, and automation ROI is hard to prove.	Prices and reports around measurable autonomous outcomes, so spend ties to economic impact.
CTO / Engineering	Agent demos are easy; production writes across systems are risky.	Provides the commerce-native runtime, policies, invariants, connectors and observability that production execution needs — so you do not rebuild the commerce execution layer yourself.
BPO / agency / 3PL	AI threatens labor-based delivery and margins.	Becomes your AI delivery engine: more client workflows per person, and a higher-value managed-operations offer, while you keep the customer relationship.
AI developer / platform	Building commerce state, rules, payments, inventory and fulfillment from scratch is expensive.	Embed iCommerce and give your agent a broad, policy-controlled commerce tool surface — commerce functionality in days rather than months.

Where things stand today

Not every part is at the same stage, and each API tab says which is which.

Part	Today
ResponseCX, Voice, Sync Server, EDI, NSR, Temporal Engine, Sandbox	Hosted and live; documented from the spec each server publishes
Computer Use	Reference published from the platform’s own app; the production host is being provisioned
Commerce Engine	Reference published from the engine’s own spec; a public host is the next step
Chat Widget, StateSet Agents	Runtimes you deploy — the widget calls your backend; the agents gateway is stateset-agents serve

Measured across production deployments: agents resolve 62% of contacts one-touch in the measured cohort, with an AI-handled CSAT of 4.82, and a typical deployment takes two to four weeks. Each figure is scoped to the deployment it was measured on — ask for the methodology before quoting it.

Pricing

The Engine is priced on outcomes, not seats, tokens or capacity: you pay when work is completed — a return resolved, a cart recovered, a lead qualified. Every plan combines an optional monthly platform fee with a per-outcome price; higher tiers trade a larger fee for a lower rate.

Plan	List price	Includes
Starter	No platform fee · $2.50 per outcome	One channel (email or chat), standard support, self-service onboarding
Growth	$5,000 / month ·$2.00 per outcome	All channels including voice, a success manager, priority SLAs, outcome reporting
Enterprise	Custom fee · from $1.50 per outcome	Volume discounts, dedicated infrastructure, custom SLAs, white-glove onboarding

An outcome is a verified, end-to-end resolution, billed per SKU — today a Triage Outcome (human hand-off, $0.50) and a **Resolved Customer Contact** ($2.00), with further SKUs confirmed per contract. Outcomes can be disputed within 14 days; anything at or above $100 requires human approval before the Engine acts. Voice, Chat Widget, Email and Computer Use are activated as modules with their own subscription. Details, entitlements and invoices are in Billing; the comparison with helpdesk AI, iPaaS and building it yourself is on Why StateSet.

Next steps

1. Create an account

   stateset.com/sign-up — no card required.

2. Make a first call

   The quickstart creates and launches a ResponseCX agent in five minutes.

3. Give your agent the Engine

   Add the MCP servers or load a skill into your harness.

4. Talk to us

   Book a demo or join the Discord.

Commerce that runs itself.