Agentest launches as an offline test suite that grades AI agents from OpenTelemetry traces
Show HN: Agentest – offline tests for AI agents from OpenTelemetry traces
AiEnsured CTO Srinivas Padmanabhuni released agentest, an open-source tool that reads an agent's existing OpenTelemetry GenAI traces and runs 165 tests on what the agent actually did, such as claiming success it did not earn, acting before verifying, following a hidden instruction, calling a disallowed tool or leaking personal data. It runs offline with no API key, network or LLM judge, so the same traces always produce the same pass, fail or n/a verdicts, and it ships native-log importers for LangGraph, OpenAI Agents SDK, AutoGen, smolagents and n8n. Installable via pip install agentest-ai (version 3.5.0), it offers a --strict CI gate with JUnit, SARIF, JSON, HTML, Markdown and CSV output, plus a no-install browser report card and Colab demo.
Source:Hacker News · AI · github.com