AI Tools Observability

Overview

groundcover ingests telemetry from AI coding tools and AI agent frameworks — Claude Code, Claude Cowork, OpenAI Codex, OpenClaw, and LangChain — so you can monitor how your team and your services use AI: cost, usage, errors, and latency.

These integrations cover two different use cases. Each ships a different telemetry signal, and that determines where the data lands in groundcover:

Integration	Signals emitted	Where it appears in groundcover
Claude Code	Logs, metrics (reference)	Logs, Metrics
Claude Cowork	Logs (events)	Logs
OpenAI Codex	Logs (events)	Logs
OpenClaw	Traces, metrics, logs	Traces, Metrics, Logs
LangChain (via LangSmith)	Traces (GenAI semconv)	Traces, AI Observability

Looking for the AI Observability page? That page surfaces GenAI traces from services that emit OpenTelemetry GenAI semantic conventions (gen_ai.* attributes) — typically AI agents you build yourself, instrumented with LangChain/LangSmith, Traceloop, or native OTel. Coding tools like Claude Code, Claude Cowork, and Codex ship tool-usage telemetry (logs, plus metrics for Claude Code), not GenAI traces, so they don't appear there.

Set up an integration

In the data sources page, search for the AI tool you want to integrate and follow the wizard.