Cut AI spend automatically.

Prompt caching, model downgrades, and on-device nudges trim the bill while your team keeps shipping. Every saving is measured against a real baseline — never a guess.

$ forg savings --summary
cache hit   61%  reused prompts
downgrades  142 calls → cheaper model
duplicates  38  served from cache
────────────────────────────
baseline    $112.93  this month
saved       $28.41  vs baseline
 

Cache reuse

Reuse identical prompt prefixes automatically. Skip the tokens you already paid for last time.

Model downgrade nudge

Route simple calls to a cheaper model when the result is equivalent. Measured, not guessed.

Duplicate detection

Spot repeated requests within a session and serve the cached result before billing the model.

Every saving measured, not estimated

FORG compares every optimised call against what it would have cost at full price. The delta is your saving — reconciled daily against the same baseline so finance can trust the number.

Savings · this month