Get started

Live OpenTelemetry spans in three steps.

Install from crates.io, point to your existing OpenTelemetry collector (or spin up the included local Grafana demo stack), and start coding.

1 · Install

Install from crates.io

Install the native binary with Cargo on Linux, macOS, or WSL2:

cargo install agent-otel-bridge

2 · Register Hooks

Connect your coding CLIs

Automatically wire non-blocking hooks for Claude Code, Codex, and Gemini CLI:

export OTEL_EXPORTER_OTLP_ENDPOINT="http://localhost:4318"
agent-otel-bridge init
agent-otel-bridge doctor

3 · Explore Live

Open your dashboards

View traces and metrics in your existing OTLP backend, or launch the local demo stack from the repo:

# Optional local Grafana + Tempo + Prometheus demo
docker compose up -d

View full documentation & Grafana dashboards on GitHub ↗ Read how non-blocking telemetry works

Frequently asked questions

Common questions before you install.

Will this add latency to my terminal when I use Claude Code or Codex?

No. The CLI hook hands off the event locally and exits in under a millisecond, with a strict fail-open guard. Network export happens asynchronously in the background.

Which observability backends are supported?

Any backend that accepts standard OpenTelemetry (`OTLP/HTTP`), including Grafana Cloud / Tempo / Prometheus, Datadog, Honeycomb, Jaeger, New Relic, and Dynatrace.

How does agent-otel-bridge complement Dendro?

agent-otel-bridge streams live telemetry while agents run so you can monitor adoption, tool behavior, and rate-limit headroom in real time. Dendro analyzes local session logs afterward for deep cost attribution, handoff bottlenecks, and rework analysis.