Fifteen hands-on demo workbooks, each showing one feature area of ModelRisk MCP on a simple, real-life use case. Every workbook is self-contained — its README sheet offers two ways in:
- FAST TRACK — one aggregated prompt that runs the whole demo end-to-end (build, simulate, analyse, report) in a single message. Paste it and watch.
- Step-by-step — the same journey as numbered steps, one capability at a time, each with the exact prompt to type to Claude and what to expect back. 5–10 minutes.
Getting started: save this folder anywhere on your PC, then once per session tell Claude where it is — "My ModelRisk demo gallery is in <your folder>. Open 01_build_model_from_brief.xlsx from there." After that, "open <file> from the demo folder" is enough.
Start with 00_START_HERE.xlsx — the index, prerequisites and colour legend.
| Workbook | Real-life scenario | Features shown |
|---|---|---|
| 01_build_model_from_brief.xlsx | Coffee-chain store expansion NPV | build_model_from_brief, plan_risk_model, run_simulation, undo |
| 02_fit_distributions.xlsx | Bakery daily demand from 365 days of history | fit_and_rank_distributions, fit_distribution_to_data, compute_distribution |
| 03_fit_copula_and_wire.xlsx | Construction input costs that spike together | fit_copula_to_data, fit_all_data_and_wire |
| 04_reverse_stress_test.xlsx | Data-centre budget breach, post-mortem in advance | build_model_from_brief, reverse_stress_test |
| 05_simulation_and_charts.xlsx | Consumer-gadget launch profit | run_simulation, create_histogram_chart, create_cdf_chart, create_tornado_chart |
| 06_audit_model.xlsx | Inherited "messy" model health-check | audit_model, diagnose_workbook, find_hard_coded_inputs, restore_cell |
| 07_insurance_aggregate.xlsx | Motor insurer annual loss + capital adequacy | create_aggregate (FFT), create_risk_event, get_tail_risk |
| 08_scenarios_pricing.xlsx | SaaS pricing decision — $79 / $99 / $119 | run_scenarios, replace_constant_with_distribution |
| 09_tail_risk_op_losses.xlsx | Bank operational-risk 99.9% capital | fit_tail (GPD), get_tail_risk, create_aggregate_mc |
| 10_time_series_and_construction.xlsx | CFO 5-year stochastic revenue forecast | create_time_series, create_copula, set_named_range, save_workbook_as |
| 11_executive_report.xlsx | Board pack for a quarry acquisition | build_executive_report, build_drivers_report, generate_executive_summary |
| 12_fit_time_series.xlsx | CFO revenue projection — dynamics fitted from history | fit_time_series, read_vmrs, set_active_vmrs |
| 13_compare_options.xlsx | Two suppliers, one contract — P(A beats B) | compare_distributions, compute_correlation_matrix |
| 14_backtest_forecast.xlsx | Was last year's forecast honestly calibrated? | backtest_output |
| 15_uncertainty_sources.xlsx | More data or more capital? Epistemic vs aleatory | decompose_uncertainty |
- Windows Excel with the ModelRisk add-in installed and licensed.
- modelrisk-mcp connected to Claude (Claude Desktop / Claude Code) — see the setup guides.
- The workbooks contain no live Vose formulas — each is a deterministic starting point (data + plan). The Vose layer is built by Claude, in front of you, following the README steps. That's the demo: the agent does the modelling work, and every write is journaled and reversible.
- Shared formatting: navy title bands, light-blue cells = inputs you can change, pale-yellow cells = where Claude writes Vose formulas / results, green = the key decision metric, grey italic = a prompt to type to Claude.
- All synthetic data is generated with fixed seeds (noted on each Data sheet), so the demos are reproducible. All files recalculate with zero formula errors in Excel.