This is a live query against the real VAAS-X memory substrate — not a mockup, not a video. It runs against a public demo device on the same production resolver every paying customer uses.
One substrate, unchanged codebase, five unrelated domains stored side by side. Predictive maintenance is one of them, not the whole story.
This query doesn't mention a pump, a market, a network, or a patient — just the shape of the pattern: a quiet deviation, still inside the alert threshold, ahead of a bad outcome. Watch what comes back.
Get a free API key — no card. Then add the connector in Claude or Cursor. Store one real day’s work. Next session, ask what worked.
Connector URL: https://mcp.vaasx.com/mcp
The number next to each result is a relevance score from VAAS-X's retrieval engine — how closely that stored episode matches what you asked, with episodes tied to meaningful outcomes weighted up rather than ranked on raw text similarity alone. It isn't a percentage or a confidence level, and it isn't capped at 1 — a strong match (especially a near-exact one) can score above it. The score bar is scaled for quick visual comparison within one set of results, not as a 0–100% gauge. Compare scores against each other within the same query; the scale isn't meant to be read against a fixed threshold or compared across different queries.
Each domain below has its own three-episode story already stored: a baseline, a precursor event quiet enough to stay under the existing alert threshold, and the outcome that followed.
Describe a new state, action, or outcome from any domain. It's written to the substrate immediately — no retraining, no schema, no per-domain setup. Query for something similar afterward and it will already know.
vaasx_public_demo, same infrastructure real customers use. Content on this page is synthetic and created for demonstration — no customer or production data. Requests are rate-limited per visitor.