om-parser

Upload the OM as a PDF or DOCX, or paste its text directly into the conversation. The Skill walks through eight standard OM sections, extracts fields like NOI, cap rate, occupancy, and rent roll, and returns a structured OMExtractedData object with gap flags.

Extractable fields include property details (address, type, year built, units, square footage), financials (asking price, NOI, cap rate, GPR, vacancy, OpEx ratio), projections (IRR, exit cap, rent growth), and a unit-level rent roll with current and market rents.

Scanned image-only PDFs must be routed through an OCR pipeline before extraction, since there is no text layer to parse. OCR quality issues are flagged in the output, and heavily degraded documents receive low confidence scores with a manual review recommendation.

Yes, all extraction works standalone. Paste the OM text, upload the PDF or DOCX, or provide key data points manually. Optional MCP connections like Nexus, Box, and Egnyte only add automation for data room access and portfolio cross-referencing.

The Skill runs internal consistency checks, verifying that GPR minus vacancy approximates EGI and EGI minus OpEx approximates NOI. It also flags out-of-range metrics for the property type and assigns a calibrated confidence score from 0.0 to 1.0 with documented drivers.

Unpopulated fields are set to null and listed in the extractionGaps array of the output. The consistencyFlags array captures internal contradictions, and the confidence assessment documents exactly which gaps are limiting extraction quality.