MCP Tool Evaluation and Testing in Agent Evals
Dynamic tool discovery breaks traditional pass-fail testing designed for fixed tool menus.
Dmitri Okonkwo
Staff Writer
Dmitri Okonkwo
Dmitri holds a background in distributed systems research and spent several years contributing to open-source observability tooling before joining The Harness. He covers the nuts-and-bolts engineering work that goes into building, testing, and shipping reliable agent infrastructure.
7 stories
Regression Suite Construction From Agent Failure Traces
Catch AI agent failures hidden in production traces before they hide in averages.
Dmitri Okonkwo
Human-in-the-Loop Review for Agent Eval Calibration
Human reviewers must anchor agent evaluators to real production behavior.
Dmitri Okonkwo
Tracing Tool Call Chains in LangChain and LangGraph Agents
Deep visibility into state changes and loop cycles reveals why LangGraph agents fail.
Dmitri Okonkwo
Building an LLM Eval Pipeline in CI/CD for Agent Changes
Production traces reveal real failure patterns that synthetic tests miss entirely.
Dmitri Okonkwo
Prompt Ambiguity Root Cause Analysis
A taxonomy reveals how to trace agent failures back to ambiguous prompts systematically.
Dmitri Okonkwo
Replay-Based Eval Against Production Traces Before Shipping Changes
Freeze production failures in test by replaying incidents with real tool outputs.
Dmitri Okonkwo