Using the calculators from an agent

If you’re building an agent or LLM app, the fastest path is the MCP server — each calculator becomes a tool. (Setup and client config: MCP server.) Prefer plain HTTP? Every tool is also a POST /v1/tools/{id} route on the HTTP API. The math is identical across both.

The tools

tdee, body-fat, one-rep-max, macros, activity-multiplier, powerlifting-attempts, muscle-potential, ffmi, rsmi. Each tool’s inputs, methods, and a worked example live on its tool page — that’s the source of truth for what to send.

Sending input

• Units are explicit. Height/weight are { "value": 180, "unit": "cm" }, never a bare number. Accepted units: cm/in for length, kg/lb for mass.
• Optional fields have defaults. e.g. tdee runs all four formulas unless you pass methods. You usually only send the required fields.

Reading the result

Every tool returns the same shape — explained in full under methods and consensus:

• results[] — one { value, unit, detail } per formula that ran.
• consensus — { mean, median, min, max, n }. Use mean/median as the headline number, min/max as the honest range.
• skipped[] — formulas that couldn’t run, each with a reason (usually a missing optional input). If a number you wanted is missing, check here first.

Over MCP you also get structuredContent (validated against the output schema) alongside a JSON text fallback.

When something is wrong

Invalid or impossible input does not crash the session — it comes back as a tool error (isError: true over MCP; an { "error": { type, message, details } } envelope over HTTP). The type tells you which: validation_error (input failed the schema), domain_error (valid but impossible, or a required method’s inputs are missing), not_found (unknown tool id).

A chained workflow

Many real questions span tools. To set a cut for a lifter, chain them:

1. body-fat → estimate body-fat % from circumferences.
2. tdee → feed that body_fat in to unlock the lean-mass methods and get maintenance calories.
3. macros → take a calorie target below maintenance and split it into protein/fat/carbs.

Each step’s output is plain JSON you carry into the next call.