← RoyalBit
Elastic License 2.0

Forge

Battle-tested financial math
for AI that can't afford to be wrong.

173 functions validated against Gnumeric & R.
MCP integration for Claude. 50x more token-efficient than Excel.

173
Functions
2,133
Tests
100%
Coverage
50x
Token Efficiency

Battle-Tested Math

Every Forge formula is validated against two independent sources.
If Gnumeric and R agree, so does Forge.

Gnumeric

714

formulas validated

Academic-grade spreadsheet engine

R Statistical

2,957

conditions verified

Gold standard for statistical computing

No trust required. Verify everything.

What Excel Can't Do

7 analytical engines that require add-ins, VBA, or external tools in Excel.
Built-in with Forge.

Monte Carlo Simulation

10,000 iterations, 6 distributions, Latin Hypercube

Bootstrap Resampling

Confidence intervals from historical data

Decision Trees

Sequential decisions with backward induction

Real Options

Black-Scholes and Binomial Tree pricing

Tornado Diagrams

Sensitivity analysis, one-at-a-time

Bayesian Networks

Causal dependency with belief updates

Scenario Analysis

Best/worst/base case comparisons

MCP Integration

Native Model Context Protocol server for Claude.
10 tools. One command: forge mcp

forge_calculate
forge_validate
forge_audit
forge_export
forge_import
forge_simulate
forge_predict
forge_scenarios
forge_functions
forge_version
# Add to Claude Desktop config
"forge": {
  "command": "forge",
  "args": ["mcp"]
}

Claude gets verified financial calculations.
No hallucinations. No wrong numbers.

YAML-Native

# budget.yaml - Financial model as code
_forge_version: "5.0.0"

assumptions:
  revenue:
    value: 1000000
  cogs:
    value: 400000
  gross_profit:
    formula: "=revenue - cogs"
  gross_margin:
    formula: "=gross_profit / revenue"

LLMs are trained on 13M+ YAML files. Formulas preserved as text.
Git-native. AI-friendly. Token-efficient.

Capabilities

πŸ“Š

173 Functions

Excel-compatible + FP&A native: NPV, IRR, XNPV, XIRR, VARIANCE, BREAKEVEN.

🎲

Monte Carlo

Normal, Triangular, Lognormal, PERT, Discrete. Latin Hypercube sampling.

🌳

Decision Trees

Sequential decisions with backward induction. Optimal path calculation.

πŸ’Ž

Real Options

Value flexibility to defer, expand, or abandon. Black-Scholes pricing.

πŸŒͺ️

Tornado Diagrams

Sensitivity analysis showing which inputs drive variance.

πŸ”

Bootstrap

Non-parametric confidence intervals. Percentile and BCa methods.

πŸ•ΈοΈ

Bayesian Networks

Causal modeling with belief updates. Evidence propagation.

πŸ”„

Excel Import/Export

Import .xlsx to YAML. Export with working formulas. Bidirectional.

πŸ€–

MCP Server

10 tools for Claude. Native integration. One command setup.

πŸ“

Git-Trackable

Diff, branch, PR, blame. Real audit trail for financial models.

⚑

Local-First

Runs on your machine. No cloud. No telemetry. Air-gapped ready.

πŸ”¬

Dual Validation

Every formula tested against Gnumeric and R. Zero tolerance for errors.

Why Forge?

Capability Excel FP&A Platforms Forge
Git version control Binary files Proprietary Native YAML
AI token efficiency ~100k tokens Cloud API ~2k tokens (50x)
MCP integration None None 10 tools native
Monte Carlo built-in Requires add-in Some vendors 6 distributions
Works offline Yes Cloud required 100% local
External validation None None Gnumeric + R

Get Started

Source available under Elastic License 2.0.
Free for evaluation, development, and non-production use.

# Install from source
cargo install royalbit-forge

# Or clone and build
git clone https://github.com/royalbit/forge
cd forge && cargo build --release

License

Elastic License 2.0 — Source available, not open source.

  • βœ“ View, fork, and modify the source code
  • βœ“ Use for evaluation and development
  • βœ“ Use for non-production purposes
  • βœ— Commercial production use requires license
  • βœ— Cannot offer as a managed service

For commercial licensing, open a GitHub issue.

Built with AI

1 human + AI. Powered by Claude from Anthropic.