spend-cap

client-side hard cap for AI API spending — no provider support needed

install

npm install @safetnsr/spend-cap

usage

# add to any Node.js app that calls AI APIs
SPEND_CAP=5 node --import @safetnsr/spend-cap/register app.js

# or with --require for CJS
SPEND_CAP=5 node -r @safetnsr/spend-cap/register app.js

when your daily spend hits $5, the next API call throws SpendCapError instead of making the request.

how it works

patches globalThis.fetch to intercept requests to:

• api.openai.com
• api.anthropic.com
• generativelanguage.googleapis.com

tracks cumulative daily spend in ~/.spend-cap-state.json. resets at midnight UTC.

cli

# check today's spend
spend-cap status

# reset spend counter
spend-cap reset

flags

| env var | description | default | |---------|------------|---------| | SPEND_CAP | daily cap in USD | (required) | | SPEND_CAP_STATE_PATH | custom state file path | ~/.spend-cap-state.json |

supported models

accurate pricing for: gpt-4o, gpt-4o-mini, gpt-4.1, gpt-5.4, claude-sonnet, claude-opus, claude-haiku, gemini-flash, gemini-pro. fallback estimation for unknown models.

agent interface (--json)

spend-cap status --json outputs:

{
  "date": "2026-03-07",
  "spentUsd": 3.42,
  "capUsd": 5.00,
  "remainingUsd": 1.58,
  "percentUsed": 68.4
}

pair with

• @safetnsr/vet — audit your AI coding workflow
• @safetnsr/pinch — track AI costs over time