mcpbench GitHub

server/elicit-url-secret

openai/gpt-5.6-luna:medium 2025-11-25 docs: full score 0% task-failure musts 1/3 · shoulds 0/0 · mays 0/0 works no · conformant no

Checkpoints

MUST (any failure zeroes the score — 1/3 passed)

Checkpoint Description Detail
secret-via-url-mode Calling connect_billing yields a url-mode elicitation with a valid URL tools/call connect_billing never issued an elicitation
no-secret-schema-in-band No elicitation for connect_billing used form mode or carried a requestedSchema
secret-flow-works The tool completed with {linked: true} after the web flow; the key never appeared in-band final result does not include "linked"

Usage & cost

Nominal cost
$0.124
Input tokens
49,008
Output tokens
4,707
Cache read
471,040
Duration
116s
Timestamp
2026-07-27 17:14:35Z
Bench version
0.3.0

Cost is nominal (public API pricing): token usage × published rates; actual marginal cost is subscription-covered.