A behavioral feature layer for graph and temporal data.

hypertopos turns relational data into a geometric coordinate space where:

• entities become positions
• relationships become structure
• behavior becomes movement
• anomalies become distance from the population

This layer sits between raw data and downstream systems:

• for exploration (analysts, agents)
• for feature generation (ML pipelines)
• for monitoring (drift, regime change)

No model training required.

Core engine. Transforms relational data into a geometric space where entities become coordinates, relationships become structure, and distance becomes signal.

Agent interface. Exposes the geometric space as MCP tools — AI agents navigate, inspect, and reason about data structure directly.

Investigation workflows. Structured behaviors that guide agents through real tasks: anomaly triage, fraud investigation, drift monitoring.

Data → geometry (hypertopos-py) → tools (hypertopos-mcp) → reasoning (hypertopos-skills)

Research-stage project. Working code, reproducible benchmarks, active development. API may change.

pip install hypertopos