QUENCHZ
deciding in the open what a tool server gets free while it is still somebody's subprocess.
A stdio MCP server gets four things free from having one caller who shares its context. Once the transport is remote each becomes a decision, and this takes all four in the open and proves them from outside the Python tree.
- Tests
- 138
- Python
- 3.11 to 3.13
- Release
- v0.1.0
- Licence
- MIT
The demo, as it really ran
$ uv run python examples/two_callers.py
QUENCHZ what a tool server has to decide once it stops being a subprocess AUDIENCE a token is only good at the door it was minted for ------------------------------------------------------------------------------ minted for this resource ACCEPTED minted for another resource refused minted for this resource AND another refused minted with no audience at all refused signed by an issuer we do not trust refused Every one of these is correctly signed, unexpired and carries the right scope. REACH a caller cannot learn what it may not use ------------------------------------------------------------------------------ tools this caller can see: ['calendar.why', 'rates.window'] tools that actually exist: ['calendar.why', 'rates.window', 'series.catalogue'] asking for one it may not use: 'no tool by that name is available to this caller' asking for one that never was: 'no tool by that name is available to this caller' identical bytes: True BUDGET one burst cannot touch another caller's reserve ------------------------------------------------------------------------------ design greedy quiet one shared bucket 60 0 a reserve each, plus a spare 45 15 The vendor publishes no rate limit, no rate-limit header, and does not list 429 among its status codes, so this budget is self-imposed and reacts to nothing. COVERAGE every answer says what it did not deliver ------------------------------------------------------------------------------ requested 12 calendar days, 2026-03-30 to 2026-04-10 expected 8 delivered 8 absent, target_closed 4 absent, not_yet_published 0 absent, no_such_observation 0 absent, before_the_series 0 complete: True Four days are missing and none of them is a gap. Good Friday, Easter Monday and the weekend are closures, and the vendor's payload says nothing about any of them. ------------------------------------------------------------------------------ Source: ECB statistics.
Output captured on 2026-08-28. It is committed to the repository and a test fails when it stops matching a live run, so this page cannot quietly drift from the code it describes.
Read it
- Source on GitHub the README carries the argument in full
- Release v0.1.0 changelog and tag
The rest of the toolset
- QUACKZ deflating a backtest that only looks good because it was picked out of two hundred
- QUOTEZ market data an agent can read and cannot act on
- QUELLZ measuring what prompt-injection containment costs in utility as well as in attack rate
- QUIDZ refusing the outbound payment that would have gone out twice
- QUESTZ stopping a scraper before it writes a CSV from a page that changed shape
- QUIZZ answering what a statistic said at the time, and refusing when it cannot
- QUARANTINEZ treating an outcome the venue never confirmed as terminal rather than as a retry
- QUILTZ proving infrastructure code wrong without a cloud account, and saying what that cannot show
- QUAYZ telling a crash loop from an OOMKill, and naming the failure that no single field finds
- QUARRYZ keeping every version a statistical office published, and failing the build when it quietly issues another
- QUASHZ refusing a row whose outcome had not been decided yet when the decision would have been made
- QUALMZ a fixed number of looks at the holdout, where re-running the same configuration does not buy another
- QUEUEZ ordering a feed by its sequence, because on a real recorded session the clock goes backwards
- QUANDARYZ counting the distinct screens a component can settle into when its responses arrive out of order
- QUIETZ watching whether the data arrived rather than whether the server answered