We instrumented Claude Code for our team. Here's what the data showed
From ad hoc prompting to spec-driven agents, plus self-hosted Claude Code telemetry with OTel, Prometheus, Loki and Grafana.
Written by
I lead engineering at Oak Network and build agentic AI systems that run in production, not in demos.
I lead engineering at Oak Network and build agentic AI systems that run in production, not in demos. 18+ years shipping software, the last several leading remote teams for US companies. My focus is agentic AI: multi-agent orchestration, LLM-powered tooling, and instrumenting agents for cost, observability and usage. • 4 AI agents in production at Oak, instrumented end to end • Agentic development culture measured by cost and usage per developer • 25-engineer team across US and LatAm • Engineering behind a protocol with 1M users • Open source: ai-workflow, spec-driven agentic development for Claude Code Stack: TypeScript, Go, Python, Rust, React/Next.js, AWS, Kubernetes, EVM/Solana. rafagomes.tech
Responses
Markdown supported · ⌘↵ to post
No responses yet.