Test certificate: buggy-test-serverhttp://127.0.0.1:4310/mcp
Failures: fix these first 6
slow_report: Has no description, so the AI has to guess what it does
fetch_notes: Description contains hidden <IMPORTANT> block. This is a known trick to make AI do things the user didn't ask for.
slow_report: No answer within the time limit
slow_report: Hung on missing required input "id"
search_users: Error message shows what looks like a database password in a link
slow_report: Hung on wrong type for "id"
Notes: worth fixing 10
4 of 9 tools have no safety label (read-only or dangerous): list_projects, team_digest, get_or_create_user, delete_account
search_users: Slowest tool: search_users (3.0 s): users will notice the wait
search_users: Inputs with no description: name
delete_account: Name sounds like it deletes something, but it is not labeled dangerous
fetch_notes: Answered, but it gives the exact same answer to bad input as to good input, so it may be ignoring what it is sent
fetch_notes: Accepted missing required input "key" without complaining
and 4 more in the test items below
All test items
Open an item to see every result behind it.
T-1ConnectsConnected in 80 msPASS
Connected in 80 ms
T-2Ready for the newest MCP versionUses an older MCP version (2025-11-25). It works today, but it isn't on the July 2026 version (2026-07-28) yet.NOTE
Uses an older MCP version (2025-11-25). It works today, but it isn't on the July 2026 version (2026-07-28) yet.
T-3Lists its toolsFound 9 toolsPASS
Found 9 tools
T-4Every tool is describedslow_report: Has no description, so the AI has to guess what it doesFAIL
search_users: Inputs with no description: name
slow_report: Has no description, so the AI has to guess what it does
T-5Tools are labeled safe or dangerous4 of 9 tools have no safety label (read-only or dangerous): list_projects, team_digest, get_or_create_user, delete_accountNOTE
4 of 9 tools have no safety label (read-only or dangerous): list_projects, team_digest, get_or_create_user, delete_account
delete_account: Name sounds like it deletes something, but it is not labeled dangerous
T-6No hidden instructionsfetch_notes: Description contains hidden <IMPORTANT> block. This is a known trick to make AI do things the user didn't ask for.FAIL
fetch_notes: Description contains hidden <IMPORTANT> block. This is a known trick to make AI do things the user didn't ask for.
T-7Tools work with good inputsslow_report: No answer within the time limitFAIL
get_weather: Worked (5 ms)
search_users: Worked (3.0 s)
fetch_notes: Answered, but it gives the exact same answer to bad input as to good input, so it may be ignoring what it is sent
slow_report: No answer within the time limit
strict_lookup: Worked (22 ms)
list_projects: Not run: no safety label (if you own this server, test again in owner mode)
team_digest: Not run: no safety label (if you own this server, test again in owner mode)
get_or_create_user: Not run: no safety label (if you own this server, test again in owner mode)
delete_account: Not run: name sounds like it deletes something
T-8Bad inputs give a clean errorslow_report: Hung on missing required input "id"FAIL
get_weather: Refused missing required input "city" with a clear error
get_weather: Refused wrong type for "city" with a clear error
search_users: Refused missing required input "name" with a clear error
search_users: Refused wrong type for "name" with a clear error
fetch_notes: Accepted missing required input "key" without complaining
fetch_notes: Accepted wrong type for "key" without complaining
slow_report: Hung on missing required input "id"
slow_report: Hung on wrong type for "id"
strict_lookup: Refused missing required input "code" with a clear error
strict_lookup: Refused wrong type for "code" with a clear error
Errors are sent with an HTTP error status (like 400) instead of a normal MCP reply, so some AI apps may treat them as crashes: strict_lookup
T-9Tools work togetherNo tool's answer could be fed into another tool, so there were no chains to test (not counted either way)NOT TESTED
No tool's answer could be fed into another tool, so there were no chains to test (not counted either way)
T-10Fast enoughsearch_users: Slowest tool: search_users (3.0 s): users will notice the waitNOTE
Connecting took 80 ms
Listing tools took 6 ms
Average tool answer: 762 ms
search_users: Slowest tool: search_users (3.0 s): users will notice the wait
T-11No secrets leakedsearch_users: Error message shows what looks like a database password in a linkFAIL
search_users: Error message shows what looks like a database password in a link
search_users: Error message shows internal details: code stack trace, server file path
Tools not run, and why
QAmcp never runs a tool that could change something on your server. If you own it, --owner runs the unlabeled tools whose names clearly only read.
- list_projectsno safety label (if you own this server, test again in owner mode)
- team_digestno safety label (if you own this server, test again in owner mode)
- get_or_create_userno safety label (if you own this server, test again in owner mode)
- delete_accountname sounds like it deletes something