Guardamos

Guardamos is a third-party content audit API for companies building AI-powered fitness, training, and wellness features.

THE PROBLEM Most AI coaching features are a prompt wrapped around an LLM. The output reads like expert advice — specific sets, reps, intensities, progressions — but nothing checks it against published clinical or sports-medicine guidance. Teams shipping these features have no defensible way to show their AI output was reviewed against anything.

HOW IT WORKS Send Guardamos an AI-generated training plan via API. It returns:

• A verdict: pass, flagged, or rejected
• The specific rules evaluated, with source attribution
• For passing plans, the confirmed facts the verdict was based on
• For flagged plans, an explanation in English, Portuguese, or Japanese

ARCHITECTURE Layer 1 is a deterministic rule engine — plain code, no LLM in the judgment path. Rules are extracted from published position statements and each is traced to a verbatim quote in the source PDF before it ships. Current coverage spans general adult resistance training (ACSM), safe return to training after inactivity (CSCCa/NSCA), overtraining prevention (ECSS/ACSM), and a dedicated rule set for older adults (NSCA).

Layer 3 generates plain-language explanations, but only from facts Layer 1 already confirmed — never a free re-reading of the plan.

WHY THIRD-PARTY MATTERS The rule engine is open source, and you can self-host it. But self-hosting produces first-party self-attestation. The hosted API is what makes the check genuinely independent — a distinction that matters as AI governance frameworks move toward requiring documented, auditable risk management.

PRICING $149/month with a 14-day free trial. Self-hosting is free (MIT licensed).