Research prototype that connects Quantrix Modeler, a multidimensional spreadsheet for financial modelling, to the Claude AI assistant through the Model Context Protocol (MCP). It has three components: a Java plugin for Quantrix Modeler (`it.unitn.lerzegov.actions` 1.0.1) with an embedded HTTP server and canvas actions for model building; an MCP server in Node.js (version 7.3.0, 40 tools); and a Claude skill for Quantrix scripting. With them the assistant reads, builds and modifies models through their own structure (matrices, categories, items, formulas): it explains how a cell is computed, writes values with guards against writes to the wrong model or to partial selections (every call can name its target model, which the server activates and verifies before running), runs parameterized library scripts that create matrices from JSON, exports tables, charts and canvases, and manages several open models. It continues the author's 2025 bridge (Groovy HTTP server and MCP server 3.0.0, DOI 10.5281/zenodo.23064178), with a new architecture: the HTTP server was rewritten in Java as a Quantrix plugin, the MCP server grew from 12 to 40 tools, and a Claude skill encodes the modelling workflow. Developed by the author between January and September 2026 and used in teaching and research: financial statement analysis, financial planning and LBO models, and the semi-automatic review of Quantrix models submitted as course assignments (in September 2026, 81 students and 1,077 model revisions).
Quantrix–Claude MCP bridge (2026): plugin per Quantrix Modeler, server Model Context Protocol e skill di Claude per la modellistica finanziaria assistita dall’intelligenza artificiale / Erzegovesi, L.. - ELETTRONICO. - (2026). [10.5281/zenodo.23078960]
Quantrix–Claude MCP bridge (2026): plugin per Quantrix Modeler, server Model Context Protocol e skill di Claude per la modellistica finanziaria assistita dall’intelligenza artificiale
Erzegovesi, Luca
2026-01-01
Abstract
Research prototype that connects Quantrix Modeler, a multidimensional spreadsheet for financial modelling, to the Claude AI assistant through the Model Context Protocol (MCP). It has three components: a Java plugin for Quantrix Modeler (`it.unitn.lerzegov.actions` 1.0.1) with an embedded HTTP server and canvas actions for model building; an MCP server in Node.js (version 7.3.0, 40 tools); and a Claude skill for Quantrix scripting. With them the assistant reads, builds and modifies models through their own structure (matrices, categories, items, formulas): it explains how a cell is computed, writes values with guards against writes to the wrong model or to partial selections (every call can name its target model, which the server activates and verifies before running), runs parameterized library scripts that create matrices from JSON, exports tables, charts and canvases, and manages several open models. It continues the author's 2025 bridge (Groovy HTTP server and MCP server 3.0.0, DOI 10.5281/zenodo.23064178), with a new architecture: the HTTP server was rewritten in Java as a Quantrix plugin, the MCP server grew from 12 to 40 tools, and a Claude skill encodes the modelling workflow. Developed by the author between January and September 2026 and used in teaching and research: financial statement analysis, financial planning and LBO models, and the semi-automatic review of Quantrix models submitted as course assignments (in September 2026, 81 students and 1,077 model revisions).| File | Dimensione | Formato | |
|---|---|---|---|
|
nota-tecnica-quantrix-mcp-bridge.pdf
accesso aperto
Descrizione: Nota tecnica
Tipologia:
Altro materiale allegato (Other attachments)
Licenza:
Altra licenza (Other type of license)
Dimensione
1.77 MB
Formato
Adobe PDF
|
1.77 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione



