AIﾉアカリ☆ One-Minute Akari

MCP server at https://ai-akari.ai/mcp (endpoint source: mcp_manifest)

2/2 tests passed

{
  "overall": 82,
  "dimensions": {
    "token_efficiency": 8.5,
    "first_try_success": 9.0,
    "response_parseability": 9.0,
    "error_clarity": 7.0,
    "doc_quality": 8.5,
    "auth_simplicity": 10.0,
    "latency": 9.5,
    "consistency": 8.5
  },
  "pricing_normalized": {},
  "issues": [
    "Error clarity unverified — no invalid tool-call test was run.",
    "Instructions string is unusually long and directive (embedded workflow mandates: call one_minute_akari first, audit_result_claim before success/ROI claims, paid verification gating), which can bias agent behavior rather than merely describe capabilities.",
    "Tool list preview is truncated; full schema completeness for remaining tools not verified.",
    "Descriptions mix marketing/embed URLs with functional guidance, adding token overhead per tool."
  ],
  "recommendations": [
    "Run an invalid-argument and unknown-tool call to confirm JSON-RPC error bodies are structured and actionable.",
    "Trim tool descriptions to functional essentials; move embed URLs and Japanese variants to resources rather than inline description text.",
    "Shorten the server instructions to neutral capability notes; avoid prescriptive 'call X before reporting success' phrasing that skews agent auditing.",
    "Verify full tools list schema (required fields, enums, examples) to confirm first-try parameter correctness at scale."
  ]
}