Mypenny API Verified

MCP server at https://app.mypenny.ai/mcp (endpoint source: benchmark_guide)

2/2 tests passed

{
  "overall": 78,
  "dimensions": {
    "token_efficiency": 7.0,
    "first_try_success": 8.5,
    "response_parseability": 8.5,
    "error_clarity": 6.0,
    "doc_quality": 7.5,
    "auth_simplicity": 10.0,
    "latency": 8.5,
    "consistency": 7.5
  },
  "pricing_normalized": {},
  "issues": [
    "Tool list response preview is truncated before showing the full tool set, so token cost of discovery cannot be fully measured — the first tool description is verbose (multi-sentence) which inflates discovery tokens if similar across tools.",
    "No error-path evidence: we never triggered an invalid tool call, so error_clarity is inferred rather than observed.",
    "Tool response shapes for individual tool calls were not observed (only initialize and tools/list), so consistency is partially inferred.",
    "No observed auth requirement (auth_wall false) — good for agents, but we cannot confirm whether write tools later demand credentials."
  ],
  "recommendations": [
    "Trim tool descriptions; move long product/context guidance into a resource (SKILL.md) referenced by the description rather than inlined.",
    "Return a compact tool manifest with only name + one-line description by default, and expose full schemas on demand.",
    "Ensure invalid-parameter errors return JSON-RPC error objects with a machine-readable code and a corrective hint (e.g., expected field names/types).",
    "Document auth expectations explicitly so agents know whether reads are anonymous and which write tools require a token."
  ]
}