CleanClean
MCP Tools

get_token_savings

Show cumulative token savings from Clean's compact result format versus raw JSON.

get_token_savings reports how many tokens Clean has saved by returning compact tiered/TOON-formatted results instead of full JSON. On every search, Clean formats the results twice — the compact summary it actually sends, and a full-JSON baseline — and banks the character-count difference (divided by four) as an estimated token saving.

Inputs

NameTypeRequiredDefaultDescription
resetbooleanfalseReset all statistics to zero instead of printing them.

Example

{ "name": "get_token_savings", "arguments": {} }

Returns a formatted report: total searches, JSON vs compact character counts, estimated tokens saved, average savings percent, and a per-session breakdown.

To clear the counters:

{ "name": "get_token_savings", "arguments": { "reset": true } }

How the number is measured

The headline figure is a concrete per-search measurement, not a marketing estimate — it's the literal character-count difference between what was sent and what a naive server would have sent (a full-JSON dump including every entity's complete source), divided by four. The local edition persists lifetime totals to ~/.clean/stats.json; session counters reset each time the server starts.

See How Clean reduces cost for the full accounting model.

On this page