Compression control

Intensity levels

Three levels let you trade compression against readability. lite is gentle, full is the sweet spot, ultra is maximum density. Set a default per API key, override per-request if needed.

The three levels

Level	Avg. savings	Feel	Use when
`lite`	~40%	Professional, natural prose. Articles preserved.	Customer-facing output, marketing copy, anything a user reads directly.
`full` (default)	~65%	Fragment style. Articles dropped. Still grammatical enough to read fast.	Internal tools, dev workflows, technical explanations, summaries for other AI.
`ultra`	~75%	Maximally abbreviated. Uses arrows (→) for causality. Almost a notation.	Logs, pipelines, agent-to-agent communication, places where no human reads the output.

Setting a default per API key

Every mintoken key has an intensity field. Set it when you create the key (in the dashboard or via POST /v1/auth/keys with intensity: "ultra") and every request authenticated with that key uses it.

Per-request override

Sometimes you want ultra for machine-to-machine traffic but lite for anything shown to a user. Override with X-Mintoken-Intensity or the ?intensity= query param:

curl https://api.mintoken.in/v1/chat/completions \
  -H "Authorization: Bearer mt_live_xxxxx" \
  -H "X-Provider-Key: sk-proj-xxxxx" \
  -H "X-Mintoken-Intensity: ultra" \
  -H "Content-Type: application/json" \
  -d '{ "model": "gpt-4o-mini", "messages": [...] }'

Pick one and stick with it at first

If you're starting out, keep your API key on fulland don't override. Once you have a week of data in analytics, look at which endpoints deliver customer-facing content and carve out a separate API key on lite for them.

Confirming what ran

Every response includes an X-Mintoken-Intensity response header with the intensity mintoken actually used for that request. If you have smart detection enabled, this tells you what the classifier picked.

← Previous

Analytics

Smart detection