Skip to main content

Thyris Agentic Commerce Platform Overview

This document explains what the Thyris Agentic Commerce platform is, what it is not, which customer-facing capabilities it contains, and how it relates to Thyris Merchant Services.

Executive Summary​

Thyris Agentic Commerce is the controlled agent orchestration layer for customer-facing commerce and service experiences. It helps customer applications route user intent to approved tools, preserve session context, return normalized responses, and provide operational visibility for agent-assisted journeys.

The platform is designed for experiences such as product discovery, checkout assistance, order support, customer service, loyalty lookup, and other approved business actions exposed through MCP-compatible tools or customer backend services.

Thyris Merchant Services provides merchant-side catalog, cart, order, procurement, webhook, and store capabilities. Thyris Agentic Commerce provides the agent experience and controlled tool-access layer that can use those merchant capabilities safely.

What It Is Not​

Thyris Agentic Commerce is not a replacement for the customer frontend, merchant ERP, PIM, commerce backend, CRM, payment provider, warehouse, or supplier system. It does not require customers to expose private business logic, database design, source code, or internal operating procedures.

Customer-facing documentation should describe supported contracts, required data, integration responsibilities, and operational expectations. It should not disclose internal deployment topology, implementation internals, private infrastructure assumptions, or product implementation details.

Platform Relationship​

Customer-Facing Capabilities​

CapabilityDescription
Realm and identity modelCustomer workspace, users, service identities, access keys, and scoped access.
Merchant and store managementIndependent merchant accounts with merchant teams, developer keys, stores, languages, and currencies.
ACP CatalogMerchant- and store-scoped product synchronization, search, carts, checkouts, orders, integrations, and built-in MCP tools.
Catalog EnrichmentStore workflows using the merchant's isolated text/image provider registry with separately measured usage.
Runtime chatA public API for customer applications to send user intent and receive normalized responses.
Agent configurationCustomer-controlled assistant identities, defaults, allowed tools, and journey selection.
Flow configurationNamed business journeys such as commerce discovery, checkout support, order tracking, or customer support.
MCP server integrationRegister approved downstream tool servers and expose their capabilities through controlled access.
Tool executionExecute approved business tools directly or through configured runtime journeys.
Session continuityPreserve context across customer interactions and support workflows.
Usage reportingReview runtime usage, transaction categories, and operational activity.
Unified MCP serviceLet compatible MCP clients access approved Engine and Catalog workflows while the same service owns MCP registration, discovery, routing, and execution.
Manager CLIInspect realm runtime resources through an embedded, allowlisted, read-only terminal.
Operational validationHealth, readiness, UAT, go-live, monitoring, and support handoff guidance.
Security controlsShort-lived bearer tokens, server-side secrets, least-privilege service identities, and customer data handling guidance.

Readiness Outcomes​

A customer is ready for Thyris Agentic Commerce when it can provide:

  • A target customer experience and approved user journeys.
  • A customer backend or frontend-backend layer that can keep credentials server-side.
  • At least one approved tool surface, MCP server, or Merchant Services capability for the agent to use.
  • Clear rules for what the assistant may answer directly, when it should ask for clarification, and when it may trigger a business action.
  • A staging environment with realistic test data and support ownership.
  • Monitoring and escalation process using session IDs, trace IDs, and agreed operational contacts.
  1. Confirm the customer journey and business actions in scope.
  2. Create or receive workspace and service credentials.
  3. Validate authentication and runtime health.
  4. Configure realm runtime AI separately from the merchant and store Catalog/Enrichment setup.
  5. Register approved downstream MCP servers or Merchant Services capabilities when external tools are required.
  6. Configure agent and flow behavior at a business-policy level.
  7. Run a small runtime chat test with a known session ID.
  8. Validate frontend rendering, Catalog actions, and tool handling.
  9. Run UAT scenarios for happy path, clarification, failed tool, timeout, and support handoff.
  10. Confirm security, credential storage, logging, and production ownership.
  11. Move to production with monitoring and rollback controls in place.