MCPulse
Shows which MCP tools agents retry, get empty results from, or never call at all.
MCPulse instruments an MCP server with two lines of code and reports first-call success rate, retries, empty-but-successful responses, and dead tools per tool — failures that normal logs record as success because no error was thrown. It never sees the actual arguments or results passing through, and lists zero runtime dependencies. In its own analysis of 4,749 public MCP servers it found concrete cases: a tool retried 2.4x on average, another returning empty 12% of the time, and one costing 480 tokens of schema per session despite never being called. It's free forever, no card or trial.
What holds up
- +Catches failures normal logs miss entirely — empty-but-successful responses and silent retries
- +Never sees your tool arguments or results, by design, not just by policy
- +Free forever, no card and no trial clock, with all ten MCP SDKs supported
- +Flags tools that cost schema tokens every session but are never actually called
Mind the limits
- −Only measures MCP tool-call health — it doesn't trace your server's internal code like APM tools do
- −Needs genuine agent traffic to be useful — a server with few real calls yields thin data
Picked here? Add the badge
Put it on your site. It links back to this page. Free for every listed product.
<a href="https://stillworks.dev/products/p/mcpulse/"><img src="https://stillworks.dev/badge/mcpulse.svg" alt="Picked by StillWorks" width="230" height="54"></a>