Intelligence
with its evidence
still attached.
NovAI Limited is a technology innovation enterprise based in Hong Kong, dedicated to building edge AI Agent ecosystems.
We create dependable systems for agents and the people accountable for their decisions—systems that keep sources, rules and operational boundaries visible from input to action.
Applied intelligence for consequential work.
NovAI works where raw information must become an accountable decision—not another opaque answer.
Our systems are built for real operating environments. They connect source data to governed interfaces, deterministic rules and reviewable outcomes, so an AI agent can move quickly without hiding what happened underneath.
From financial disclosures to ocean heat stress, the domain changes but the standard does not: preserve provenance, state uncertainty honestly, expose limits and design the final mile for action.
Trust is an engineering property.
Four core architectural principles govern every system, API and agent we engineer.
Start with the source.
Results retain provenance, effective time and the context required to understand where a fact came from.
Make critical rules explicit.
Permissions, licenses, thresholds and output policies remain inspectable instead of disappearing inside a model.
Design for the operating edge.
We meet agents and teams where decisions happen: interfaces, alerts, tools, audit trails and workflows.
Keep the boundary visible.
A dependable system knows when to refuse, when evidence is insufficient and when a person must review.
argus
Agent-Readable Financial Data Layer
Argus gives AI systems a governed path from a financial question to structured, source-backed facts. It preserves the evidence, time, quality, license and restrictions that determine whether a result is safe to use.
Discover
Read the governed tool registry and lifecycle metadata.
Request
Call through CLI, REST, GraphQL or MCP.
Govern
Apply permission, license, evidence and output-policy checks.
Validate
Inspect the DataPackage and audit identity before automation.
DataPackage / governed response
Transport success is not proof that a fact is usable.- Fact
- declared value + unit + entity
- Evidence
- source document + exact provenance
- Time
- effective time + known time + as-of context
- Quality
- freshness + confidence + review state
- Rights
- license + restrictions + allowed use
- Audit
- policy outcome + request identity
Facts, not financial advice.
Argus does not provide stock recommendations, trading signals, target prices, portfolio allocation or automated trading. Refusals and restrictions remain structured, visible and machine-readable.
One shared governance boundary keeps fact, evidence, time, quality, license, restriction and audit semantics consistent across every interface.
Explore argusfa.comReefBeacon
Coral reef bleaching and marine heatwave early warning for Hong Kong.
ReefBeacon reflects a wider NovAI commitment: applying dependable intelligence beyond commercial infrastructure to the natural systems that sustain us.
Visit ReefBeacon- 9
- marine parks and reserves monitored
- Daily
- NOAA 5 km satellite SST updates
- 84
- reef-building coral species in Hong Kong
Read ocean heat stress daily.
NOAA Coral Reef Watch SST, HotSpot and Degree Heating Week products establish the environmental signal.
Keep the threshold deterministic.
Code applies explicit DHW rules: above 4 °C-weeks is Watch; above 8 is Bleaching Risk.
Move evidence into action.
Structured alerts carry the raw snapshot, threshold version, delivery record and audit trail.
Build longer ecological memory.
Survey imagery, cited historical Q&A and trend reporting connect acute events to annual insight.
ReefBeacon is an early-warning system, not real-time bleaching confirmation. Deterministic code handles numeric thresholds; AI supports orchestration and narrative, with low-confidence imagery routed for human review.
Bring dependable intelligence to the decisions that matter.
Talk to us about agent-readable data, governed AI workflows, environmental intelligence or a new collaboration.