DriftGuard (MCP pinning): figures & benchmarks

Evidence for DriftGuard (MCP pinning): 1. Evidence pack · 2. Answers to buyer questions · 3. Figures & benchmarks · 4. Selling tools · IHVRK kit · Learn it · why these pages exist

Every number on this page is read from the signed transcript inside the IHVRK kit; none is typed in by hand.

5/5

claims matched when recorded

attack cases refused or flagged as claimed

226 ms

median round trip (min 174, max 4585), incl. network

#	Kind	Claim	Recorded value	HTTP	ms
C01	control	Same definition as pinned -> No change, no alarm.	`false`	200	4585
C02	attack	Description quietly adds a BCC -> The classic MCP rug-pull.	`true`	200	204
C03	attack	Schema gains a 'bcc' field -> A new parameter is a new capability.	`true`	200	174
C04	attack	An invisible character added -> Zero-width changes are changes.	`true`	200	226
C05	attack	Screen a poisoned definition -> Poisoning is refused before it is ever pinned.	`"refused until a person acknowledges the findings"`	200	257

Copy the table (paste into a spreadsheet)

id	kind	claim	recorded_value	expected	http	ms	recorded_at
C01	control	Same definition as pinned -> No change, no alarm.	false	[false]	200	4585	2026-11-03T22:46:26+00:00
C02	attack	Description quietly adds a BCC -> The classic MCP rug-pull.	true	[true]	200	204	2026-11-03T22:46:26+00:00
C03	attack	Schema gains a 'bcc' field -> A new parameter is a new capability.	true	[true]	200	174	2026-11-03T22:46:27+00:00
C04	attack	An invisible character added -> Zero-width changes are changes.	true	[true]	200	226	2026-11-03T22:46:28+00:00
C05	attack	Screen a poisoned definition -> Poisoning is refused before it is ever pinned.	"refused until a person acknowledges the findings"	["refused until a person acknowledges the findings"]	200	257	2026-11-03T22:46:29+00:00

Download: figures.tsv · transcript.json (full answers) · whole kit

How to read these numbers

Recorded 2026-11-03T22:46:29+00:00 against cainstudio.online from our build host; times include the internet round trip.
They show these inputs, not every possible input. Known misses are on the limits page.
This is not a load test, an SLA or a third-party benchmark. Re-run it yourself: python3 verify.py --live.