Read LLM traces and coding-agent sessions from LangWatch using the langwatch CLI.
npx skills add https://github.com/langwatch/langwatch --skill langwatch
Please help me install this Agent Skill. Skill: langwatch Source: https://github.com/langwatch/langwatch/tree/main/plugins/langwatch/skills/langwatch Command: npx skills add https://github.com/langwatch/langwatch --skill langwatch
When an LLM agent, prompt, or model call misbehaves in production, you need to see what it actually did rather than guess. This Skill lets you query LangWatch traces and coding-agent sessions directly from a coding session using the langwatch CLI, so you can debug failed or slow LLM runs and look up traces by id.
langwatch trace search -q "checkout agent" to find the run, then langwatch trace transcript <traceId> to see exactly what the agent did step by step.Ask the assistant to check that the langwatch CLI is installed and signed in, then search recent LangWatch traces for the agent run you want to inspect.
High-intent search queries and answers about installing and using this skill.
Run `langwatch trace search -q "your query" --limit 10 -o json` to search traces from the last 24 hours. Widen the window with `--start-date` and `--end-date`, or filter by origin with `--origin application,evaluation,simulation`.
Use `langwatch trace transcript <traceId>` to get the readable, ordered transcript of what the agent did. Add `-o json` if you want to parse the result programmatically instead of reading it.
Run `langwatch whoami` to print the identity persisted by `langwatch login`. If it reports nobody, run `langwatch login` and complete the sign-in flow in the browser; credentials should never be guessed.
No. The `langwatch ingest hook` command reports only repository identity such as the repository, branch, and worktree the session is working in. It does not transmit file contents.
Failures differ by cause: a missing trace id, an expired session, or a self-hosted endpoint the CLI was never pointed at each produce distinct errors. Read the CLI's error message to identify the cause instead of retrying the failed read in a loop.
truefoundryInstrument Python and TypeScript applications with OpenTelemetry-based tracing via Traceloop SDK.
KappTech88Diagnose LLM behavior across tracing, evaluation, and monitoring workflows in LangSmith.
CodeAlive-AIQuery Grafana Cloud logs, metrics, traces, dashboards, alerts, and incidents for troubleshooting.
hrntknrStart a local OTLP server and query traces, logs, and metrics.