Generic APM tools couldn't tell me why my MCP tool calls were failing, so I built something that actually understands the protocol

Описание и применение

mcpobs преобразует общие HTTP-отслеживания в подробные диагностические данные для MCP, классифицируя ошибки и отслеживая прогресс длительных вызовов.

Задача

APM tools treated MCP tool calls as generic HTTP, providing only latency and status, making it impossible to diagnose failures such as missing tools, argument validation errors, or expected 401 auth handshakes.

Результат

mcpobs transforms generic HTTP tracing into meaningful MCP-specific diagnostics, improving developer productivity and reducing debugging time.

Рабочий процесс

• Install mcpobs via pip: pip install mcpobs
• Import and instrument your MCP client: from mcpobs import instrument; instrument(mcp)
• Run your application; mcpobs automatically classifies failures and tracks progress

Доказательства

Drop it into your server (pip install mcpobs, then instrument(mcp)) and it classifies failures into categories that map to an actual fix, tracks progress on long-running tools while they're still running, and shows the downstream db/http/llm calls in the same trace.

Источник