10 servers curated

Observability and Debugging Tools for AI Agents

Monitoring and debugging agentic workflows is notoriously difficult due to the non-deterministic nature of LLM tool calls and multi-step reasoning chains. Developers often struggle with "black box" execution, where identifying the root cause of a failed tool call or a latency spike requires manually parsing fragmented logs across disparate systems.

Model Context Protocol (MCP) servers bridge this gap by exposing observability data directly to your AI agent. Instead of switching contexts to a dashboard, you can now query traces, logs, and metrics using natural language. This allows agents to self-diagnose issues, analyze performance bottlenecks, and verify tool outputs in real-time within the development environment.

When selecting an MCP server, prioritize those that offer deep integration with your existing telemetry stack. Look for servers that provide structured access to traces and error logs, as these are critical for debugging complex agent interactions. Ensure the server supports the specific query languages or APIs your infrastructure relies on to maintain a seamless feedback loop.

Also Worth Trying

Langfuse MCP Java

1 stars

A Java/Spring AI native server that provides read-only access to Langfuse observability data. It is ideal for teams needing to track traces, exceptions, and prompt performance through tools like fetch_traces and find_exceptions.

8 toolsLog-LogN

Seq MCP

1 stars

Seq MCP provides controlled read access to Datalust Seq instances for deep log analysis. It auto-generates tools for official Seq API routes, making it easy to perform connectivity diagnostics and search events directly.

6 toolsMCLifeLeader

Dynatrace Managed

18 stars

Designed for self-hosted Dynatrace environments, this server enables natural language querying of problems, logs, and SLOs. It supports multi-environment configurations, making it a robust choice for complex enterprise setups.

3 toolsdynatrace-oss

OpenObserve Community

7 stars

This read-only server connects to OpenObserve via REST API, providing access to log streams and dashboard discovery. It is a great choice for teams using the community edition who need to search logs and retrieve traces efficiently.

8 toolsalilxxey

Iris Eval

5 stars

Iris Eval focuses on the quality of agent output, providing hierarchical trace logging with built-in evaluation rules. Use log_trace and evaluate_output to monitor token usage, costs, and output safety in real-time.

3 toolsiris-eval

MCP Browser Logger

0 stars

This server captures browser console logs and network requests via the Chrome DevTools Protocol. It is indispensable for debugging web-based agents, allowing you to evaluate JavaScript and inspect network traffic directly.

7 toolstao-Lionel

Rybbit Analytics

2 stars

Rybbit Analytics provides deep insights into sessions, events, and metrics. With tools like rybbit_get_metric and rybbit_list_sessions, you can query complex data across 22 dimensions to understand agent performance and user interactions.

8 toolsnks-hub

Side-by-Side Comparison

ServerStarsToolsTransportAuthor
1MCP Monitor20stdioPartha-SUST16
2Uptrace MCP Server21httpdimonb
3New Relic MCP Server16httpxelber
4Langfuse MCP Java18httpLog-LogN
5Seq MCP16stdioMCLifeLeader
6Dynatrace Managed183stdiodynatrace-oss
7OpenObserve Community78httpalilxxey
8Iris Eval53stdioiris-eval
9MCP Browser Logger07stdiotao-Lionel
10Rybbit Analytics28stdionks-hub

Keep the winning workflow in memory

Find the right server here, then save the docs, prompts, and setup rules in Conare so your agent can reuse them across clients.

Open Conare