AI project - 02
One shared context layer for every AI tool a designer or a design team builds.
Status
Working pilot
stack
Node.js, MCP, CLI, RAG, JSON Schema
type
AI Context Engineering - Personal IP
the problem
UX evidence lives everywhere — interviews, tickets, transcripts, analytics, docs — and every new AI conversation starts from zero, re-explaining who the users are and what the team already knows. That's hours lost before any real thinking begins.
What is it
A context layer built for designers, not developers. It gathers scattered UX evidence from wherever it lives, understands it against a shared model of user needs and business context, and packages it into focused, ready-to-use bundles that any AI tool or agent can pick up instantly — via an MCP server, a CLI, or a published JSON Schema.
built for designers building their own ai systems
Designers are increasingly building their own AI agents and pipelines — not just using someone else's tools. UUC Engine is the shared context layer underneath all of it: gather the evidence once, and every agent, prompt, and pipeline a team builds afterward starts from the same, always-current understanding.
The shape - brief architecture
output, illustrative example
{
"connect": {
"connectorId": "blueoil-brd::::none",
"reused": true,
"ingested": 167,
"skipped": 0
},
"pack": {
"packTemplateId": "problem-framing-pack",
"packName": "Problem Framing Pack",
"phase": "define",
"generatedAt": "2026-08-18T03:27:07.210Z",
"tokenBudget": 3000,
"tokenEstimate": 792,
"truncated": false,
"entities": [
{
"id": "pain_point:blueoil-brd::::none:FR-EA-04",
"title": "Uploaded evidence files shall be stored in secure cloud storage with access restricted to authorized users (Facility Manager, Unit Head, Sustainability Manager, Auditors).",
"tags": [
"access",
"unauthorized",
"users",
"cannot",
"attachment"
],
"createdAt": "2026-08-18T03:27:07.202Z",
"updatedAt": "2026-08-18T03:27:07.202Z",
"metadata": {},
"confidence": {
"score": 0.5,
"sourceCount": 1,
"recency": "2026-08-18T03:27:07.202Z",
"rationale": "Computed by the default heuristic enrichment/classification providers. Replace with an LLM-backed provider for higher-fidelity scoring."
},
"sources": [
{
"sourceId": "blueoil-brd::::none:FR-EA-04",
"sourceType": "blueoil-brd",
"connectorId": "blueoil-brd::::none",
"retrievedAt": "2026-08-18T03:27:07.202Z"
}
],
"relatedEntityIds": [],
"entityType": "pain_point",
"statement": "Unauthorized users cannot access attachment files; access attempts are logged.",
"severity": "medium"
}
}
}
Sample output JSON of a problem-framing pack generated on given client's BRD.
early signal
~ 70%+
faster context preparation for new AI-assisted deliverables, based on early internal use. Core engine and agent hand-off layer are built and tested; a hosted, team-ready version is next.
relevant tags
AI context engineering
knowledge system
agent tooling (MCP)
UX research ops
ux deliverables