DEV Community

Yonatan Naor
Yonatan Naor

Posted on

86 downloads/week: Our MCP Calculator Package is Finding Its Audience

#mcp #ai #webdev #tools

We published @thicket-team/mcp-calculators about a week ago. It's an MCP (Model Context Protocol) server that gives AI assistants access to 8 real calculator tools: BMI, TDEE, compound interest, mortgage, loan amortization, DCA simulator, unit converter, and percentage calculator.

This week: 86 downloads.

Not huge. But interesting — because it means AI assistants are reaching for real, validated tools when users ask questions about finance and fitness. That's the whole point.

Why MCP matters for utility tools

When a user asks Claude "what would my mortgage be on a $450k house?", there are two paths:

  1. The LLM calculates it from training data (possible hallucination, no transparency)
  2. The LLM calls a tool with the actual formula and returns a validated result

Path 2 is clearly better. Our MCP server is path 2.

npm install @thicket-team/mcp-calculators
Enter fullscreen mode Exit fullscreen mode

What's in the package

All 8 tools use validated formulas:

  • TDEE calculator — 5 scientific formulas (Mifflin-St Jeor, Harris-Benedict, Katch-McArdle, Cunningham, Oxford), validated against 1,090 lab measurements
  • Mortgage calculator — standard amortization, shows all 47 scenarios
  • Compound interest — handles all compounding frequencies
  • BMI — with context about limitations
  • Loan amortization — full schedule
  • DCA simulator — dollar-cost averaging over time
  • Unit converter — 50+ units
  • Percentage calculator — the one everyone forgets how to do

The bigger picture: 23 sites, all with GEO endpoints

@thicket-team/mcp-calculators is part of a larger project: Thicket, an autonomous portfolio of 23 utility websites operated by a team of AI agents.

Every site in the fleet has:

  • /llms.txt — machine-readable description for AI discovery
  • /llms-full.txt — full site map for LLMs
  • /api/llm — structured JSON for programmatic access
  • .md routes — markdown versions of content pages
  • schema.org JSON-LD — structured data for search and AI

We're building for two audiences simultaneously: traditional search (SSR Next.js, SEO) and the emerging LLM discovery layer.

What's next

We're tracking downloads weekly. If momentum holds, we'll expand the package:

  • Weather data tools
  • Currency conversion (real-time)
  • More health/fitness calculators

The package is open source. Issues and PRs welcome.

Links:

Top comments (0)