What is MCP sampling?

MCP servers use it to delegate LLM calls back to clients, since servers lack API keys. Essential for tools needing generation, like summarizers.

How do you trace MCP server LLM calls?

Wrap ctx.mcpReq.requestSampling() in traceSampling({model: 'gpt-4o'}). Middleware handles the rest — spans nest under tools.

Does this dashboard work out-of-box?

Yes, Prometheus queries provided. Plug into Grafana for tables, charts, alerts on rates, durations, errors.

☁️ Cloud & Infrastructure

MCP Servers Now Trace Their Own LLM Calls – No More Blind Spots in Agent Tools

Imagine debugging an AI agent where 90% of your tool's delay hides in an untraceable LLM call. This fix changes that for MCP servers, handing devs real observability.

DevTools Feed Apr 04, 2026 4 min read

Read in: Deutsch English Español Français Italiano 日本語 한국어 Português (BR) Русский Türkçe

Nested OpenTelemetry trace showing MCP tool call with inner sampling LLM span

⚡ Key Takeaways

MCP sampling calls now span fully, revealing 80%+ hidden LLM latency in tools. 𝕏
Dashboard delivers glanceable metrics: rates, P95s, errors by tool — optimize fast. 𝕏
This mirrors early microservices tracing; poised to standardize before agent swarms hit prod. 𝕏

Published by

DevTools Feed

Ship faster. Build smarter.

#LLM tracing #agent observability #mcp-servers #toad-eye

Worth sharing?

Get the best Developer Tools stories of the week in your inbox — no noise, no spam.

Originally reported by dev.to

⚡ Key Takeaways

The 60-Second TL;DR

DevTools Feed

Share this article

Worth sharing?

Related Stories

MCP's Tool Permissions Wake-Up Call: Stop Handing Agents the Keys to Everything

MCP's Prompt Injection Plague: Unchecked Tools, Massive Risks

MCP Servers: Erasing Script Hell and REST's Reign in Enterprise Glue

AI Agent Atlas Ships Dev Tools Biz, Targets $12K MRR Solo

Stay in the loop