document-intelligence

soapbox-document-intelligence

Document Intelligence for real estate ESG — one pipeline for every document that enters a portfolio: classify → extract typed fields → write to the asset → RAG-index.

Covers utility bills, energy audits, leases, PCAs, rent rolls, OMs, drawings, and regulatory filings, with per-type extraction schemas and provenance on every value (page + verbatim source text). Chart, graph, and table digitization is built in via the bundled pdf-chart-parser MCP server (this plugin absorbed the former standalone pdf-chart-parser plugin).

Design

Skill-first: classification and typed extraction run on the agent's native document vision, against strict schemas in skills/document-intelligence/references/. The one piece of infrastructure is the chart/table digitization MCP (https://pdf-chart-parser.mcp.soapbox.build/mcp), bundled in the plugin manifest.

Future work (per the 2026-06-26 ESG plugin architecture spec): a dedicated batch extraction MCP for high-volume server-side ingestion, green lease scoring, and multi-language documents.

Install

• Soapbox: soapbox install document-intelligence
• Claude Code / opencode: add this repo as a plugin (skills + bundled MCP in the manifest)

Contents

• skills/document-intelligence/SKILL.md — the pipeline
• skills/document-intelligence/references/extraction-schemas.md — per-type schemas
• commands/extract-charts.md — chart/table extraction command