Skip to content

AI Integration

shortfx is designed as a deterministic tool layer for AI agents. LLMs are great at reasoning but can produce inconsistent results for calculations, date arithmetic, or string transformations. shortfx provides tested, reliable functions that always return the same output for the same input.

Two Discovery Methods

Method Best For How It Works
MCP Server AI agents with tool-calling (Claude, GPT, etc.) A live server exposing search + execute meta-tools over the Model Context Protocol
llms.txt LLMs with file access (Copilot, Cursor, etc.) A static index of all functions with signatures and descriptions

Why Dynamic Tool Definition?

shortfx exposes 3,000+ functions — far too many to register each as an individual MCP tool. Instead, the MCP server uses a dynamic discovery pattern with a small set of meta-tools:

Meta-Tool Purpose
search_shortfx_tools Find functions by natural-language query
list_shortfx_tools Browse all functions, optionally filtered by module
get_shortfx_tool_details Get full parameter schema for a specific function
call_shortfx Execute any function by its qualified name
scientific_calculate Evaluate math expressions (AST-based, no eval) 
1. SEARCH   → search_shortfx_tools("calculate days between two dates")
2. INSPECT  → get_shortfx_tool_details("fxDate.date_operations.calculate_days_between_dates")
3. EXECUTE  → call_shortfx("fxDate.date_operations.calculate_days_between_dates",
                             '{"start_date": "2025-01-01", "end_date": "2025-12-31"}')