LangWatch demos auto‑instrumenting skills
Rogerio Chaves from LangWatch demoed 'skills' that auto‑instrument codebases to trace LLM/tool calls, surface latency and failures, and auto‑generate tests — a direct attempt to shrink the debugging gap for coding agents. (x.com)
LangWatch publishes an installable "skills" package on GitHub that lets agents add tracing with a single command (for example: npx skills add langwatch/skills/tracing). (github.com)) The skills repo includes higher‑level "recipes" such as debug‑instrumentation and improve‑setup to automate debug instrumentation and evaluation scaffolding within agent codebases. (github.com)) A companion demo repository shows a Mastra-built weather agent instrumented with LangWatch for observability, prompt management, automated evaluations, and scenario testing. (github.com)) LangWatch docs describe an OpenClaw integration that ships a built‑in OpenTelemetry exporter called diagnostics-otel to send structured traces to LangWatch without custom instrumentation. (langwatch.ai)) The platform advertises end‑to‑end features including agent simulations, turning production traces into evals, and scenario testing to compare prompts, models, and end‑to‑end behavior. (langwatch.ai)) Rogério Chaves, LangWatch co‑founder and CTO, has demonstrated these observability and agent‑testing workflows in public demos and an Open‑Source Spotlight interview and in a session on automating evals and scenarios. (youtube.com)) LangWatch announced venture backing with a €1M pre‑seed round led by Passion Capital, filed publicly on February 25, 2025. (passioncapital.com))
