Copied. Paste it into Claude Code, Cursor or Codex:Fetch and follow the instructions at https://qamcp.dev/setup.md to test my MCP server.
Your AI finds your server in your project, tests it with QAmcp, explains what's wrong in plain words, fixes it with your OK, and tests again. Read the instructions
On their server. What you send to it goes to them.
What you need
Nothing: it answered without a login
Tools
9 tools
Its own safety labels
6 read-only, 1 labeled as deleting or destroying, 2 labeled as changing things
Memory it takes
About 3,300 tokens of your AI's memory, just for its tool descriptions
Largest answer
About 124.1 KB in our test
Speed
Average answer in our test: 93 ms
Tokens are an estimate: the characters in its tool names, descriptions and inputs, divided by 4. The exact number depends on your AI model. Safety labels are what the server says about itself. From the daily test on 15 Nov 2026.
Is ai·rete·rag MCP server safe?
None of the hidden-instruction wordings we know were found in its 9 tool descriptions.
Its error messages showed no secrets or internal details.
6 of its 9 tools are labeled read-only; the others may change or send things.
Overall it scored 84/100 on 15 Nov 2026.
A test can only see what the server shows from outside, not what it does with your data behind the scenes. Connect servers from makers you trust.
What this test does not check
Its source code and packages.
What a tool quietly does behind the scenes, such as writing a file or calling another service.
Instructions hidden inside live tool answers. We only look for those in tool descriptions.
Hidden instructions worded in a way our list doesn't know yet.
Tools that could change something. Those are read, never run.
!decide: Inputs with no description: domain, query, facts, unstructured_text, response_mode, filter_retrieval_with_rules, extract_from_retrieval, fact_sources, as_of
!list_rules: Inputs with no description: domain
!get_parameters: Inputs with no description: domain
!ingest_text: Inputs with no description: domain, text, source, effective_from, effective_until, supersedes
!list_documents: Inputs with no description: domain
!get_rule_source: Inputs with no description: domain
!put_rules: Inputs with no description: domain, rules_yaml, dry_run
!import_policy_rules: Inputs with no description: domain, policy_text
note
!
Tools work with good inputs1 warning
!list_rules: Answered with about 124 KB of text. Many AI apps cut off tool answers this big, so the AI may never see the end of it.
–decide: Not run: labeled as changing data
–ingest_text: Not run: labeled as changing data
–put_rules: Not run: labeled dangerous
–import_policy_rules: Not run: labeled read-only, but its name starts with "import", which usually means it changes something. Skipped to be safe: if it only reads, a clearer name helps AI apps too, or test again with --owner
and 4 more
note
!
Follows MCP best practices (not scored yet)2 tips (not scored)
!3 tools not labeled read-only don't say whether running them twice is safe (idempotentHint): decide, ingest_text, put_rules
!9 tools don't say whether they reach outside services like the web (openWorldHint): decide, list_rules, get_parameters, ingest_text, list_documents, get_rule_source and 3 more
note
!
Resources and prompts work2 warnings (1 prompt)
!decide_case: Inputs with no description: domain, case
!decide_case: Leaving out a required input gets an internal error instead of an "invalid input" error: Internal server error
and 1 more
note
✓
ConnectsConnected in 219 ms
pass
✓
Ready for the newest MCP versionUses the latest MCP version (2026-07-28)
pass
✓
Lists its toolsFound 9 tools
pass
✓
Tools are labeled safe or dangerousEvery tool has a safety label
pass
✓
No known hidden instructionsNone of the hidden-instruction wordings we know were found in tool descriptions. New wordings can get past this check.
pass
✓
Bad inputs give a clean errorAll good
✓get_parameters: Refused missing required input "domain" with a clear error
✓get_parameters: Refused wrong type for "domain" with a clear error
✓list_documents: Refused missing required input "domain" with a clear error
✓list_documents: Refused wrong type for "domain" with a clear error
✓get_rule_source: Refused missing required input "domain" with a clear error
✓get_rule_source: Refused wrong type for "domain" with a clear error
pass
✓
Fast enoughAll good
✓Connecting took 219 ms
✓Listing tools took 192 ms
✓Average tool answer: 93 ms
✓list_rules: Slowest tool: list_rules (240 ms)
pass
✓
No secrets leakedChecked 6 error messages: no secrets or internal details
pass
✓
No known hidden instructions in tool answers (not scored yet)Read 11 answers from 5 tools: none of the hidden-instruction wordings we know were found. New wordings can get past this check.
pass
–
Tools work togetherNo tool's answer could be fed into another tool, so there were no chains to test (not counted either way)
not tested
Show this score on GitHub
Free for every server in the directory. Paste this into your README: the badge always shows the newest score, from the daily test, a test on our website or a Monitor test.