ucon
Dimensional safety for LLMs.
Unit-aware calculations that never lie.
Give AI agents numeric type safety. Chemistry stoichiometry, medical dosing, engineering calculations—ucon catches unit errors before they become disasters.
# LLM asks ucon-mcp to compute a dosage
from ucon import units, Scale
kilogram = Scale.kilo * units.gram
milligram = Scale.milli * units.gram
patient_weight = kilogram(70)
dose_rate = milligram(5) / kilogram
dose = patient_weight * dose_rate
print(dose) # <350 mg> ✓ dimensionally valid
# But this? Caught immediately:
patient_weight + dose # TypeError: incompatible dimensions
from ucon import units, Scale
kilogram = Scale.kilo * units.gram
milligram = Scale.milli * units.gram
patient_weight = kilogram(70)
dose_rate = milligram(5) / kilogram
dose = patient_weight * dose_rate
print(dose) # <350 mg> ✓ dimensionally valid
# But this? Caught immediately:
patient_weight + dose # TypeError: incompatible dimensions
Chemistry-first
Stoichiometry, molarity, dilutions—the domain where unit errors hurt most.
MCP Native
Works with Claude, Cursor, and any MCP-compatible AI client out of the box.
Provenance
Step-by-step unit algebra shows exactly how every answer was derived.
Coming Soon
Hosted ucon-mcp service
API access to dimensionally-safe calculations. Built for AI agents, educators, and anyone who can't afford unit errors.