Bring product knowledge
Start with websites, PDFs, pCon Catalog, pCon / OFML, a product conversation, existing CAD, or manual entry.
Product workspace
ConfigAI gives product teams one reviewed structure for what a product is, where its facts came from, how it can vary, which 3D model belongs to it, and which rules are allowed to act on it.
Product definition
Product type
Seating / lounge chair
Family
Leya
Source bindings
4 verified sources
3D model
Attached GLB revision
The core loop
The workspace is designed around a simple contract: agents can discover, normalize, and propose. The product system only changes after the proposal is reviewed against its sources.
Start with websites, PDFs, pCon Catalog, pCon / OFML, a product conversation, existing CAD, or manual entry.
Identify product families, products, fields, components, relationships, source identities, and the evidence supporting each claim.
ConfigAI keeps unresolved information explicit and asks targeted questions when the answer changes the product type, field model, compatibility, or revision.
Product definition
Product Types and stable field identities let imports, manual editing, rules, 3D, and agents work against the same structure instead of inventing a new schema each time.
Sources
website · PDF · pCon · OFML · conversation
Discovery
evidence capture · identity · relationships
Questions
targeted clarification only where structure changes
Product definition
type · fields · products · revisions
Execution
3D · rules · MCP · applications
Stable semantic definition
Identity
product_id · source identity · family
Configuration
width · upholstery · base · arm type
Relationships
collection · components · compatibility
Assets
images · source documents · 3D revisions
Rule bindings
stable fields referenced by exact revision
Human control
Agents and imports produce proposals. Reviewers see the product type, field changes, source evidence, identity matches, assumptions, and revision impact before the product record changes.
Agent-authored or deterministic import output is stored as a proposal tied to exact source evidence.
People accept, correct, reject, or request another revision without losing the previous proposal.
Approved changes create durable product/profile revisions and preserve the older revision for reopening.
One definition, many consumers
Once a product is reviewed, the same revision can feed the 3D workspace, rule engine, configurators, MCP tools, and future commercial or production layers.
3D Studio
Source-backed modeling, checkpoints, review, and product attach.
Rule Engine
Product decisions and assembly logic on stable field revisions.
Agent tools
Workspace agents operate against reviewed product definitions.
MCP & API
Expose product context and permitted operations to applications and agents.
Next layers
Pricing, BOM, production outputs, and a richer USD digital twin belong after the reviewed product definition, not in a second data model. They are roadmap layers, not claims that the current product flow already completes them.
Pricing & quoting
Commercial rules and price outputs on the reviewed configuration.
BOM & production
Derived components and manufacturing outputs from exact product revisions.
USD digital twin
A richer scene representation where approved variant changes can be reflected in the digital twin.
ConfigAI workspace
Start with the information your company already has. ConfigAI turns it into a reviewed product definition that every later workflow can share.