● Beta — KubeEra is in active development. Profiles are live; AI-perception measurement is rolling out.
KubeEra
O

OpenLLMetry

Observability
Open source · Apache-2.0

OpenLLMetry is an open-source observability layer for LLM applications, built as a set of OpenTelemetry instrumentations. It auto-instruments calls to LLM providers (OpenAI, Anthropic, Cohere, etc.

updated 2026-07-01

What it is

OpenLLMetry is an open-source observability layer for LLM applications, built as a set of OpenTelemetry instrumentations. It auto-instruments calls to LLM providers (OpenAI, Anthropic, Cohere, etc.), vector databases (Pinecone, Chroma, Qdrant), and frameworks (LangChain, LlamaIndex, Haystack), emitting standard OTel traces and spans you can ship to any backend that speaks the protocol — Datadog, Honeycomb, Grafana Tempo, or Traceloop's own hosted platform. The pitch is simple: don't build a proprietary LLM tracing format, extend the observability standard everyone already uses.

Who builds it and why

Maintained by Traceloop, a company selling a commercial LLM observability/evaluation platform. OpenLLMetry is their open-core wedge — free instrumentation that funnels naturally into their paid backend, though the OTel-native design means you're not locked in if you skip it. 109 contributors on GitHub suggests real community involvement beyond the core team, not just a corporate side project with token external commits.

Production readiness signal

7,253 GitHub stars, active commit history through mid-2026, and a mature release number (0.62.1) point to a project past the "weekend hack" stage and into steady iteration — the 0.x versioning after 60+ releases is more about SemVer discipline than instability. Apache-2.0 licensing removes copyleft concerns for commercial use. CNCF maturity level is not publicly available (it's not clear this is a CNCF project at all — no confirmation either way). Founding date isn't published, but the instrumentation coverage (multiple LLM providers, multiple frameworks, multiple vector DBs) suggests it's been developed over at least a couple of years, not a recent bolt-on. No public SLA, no published production case studies with company names attached — you're evaluating this on repo signals, not testimonials.

Who should use this

  • Teams already running OpenTelemetry pipelines who want LLM spans to show up alongside existing service traces instead of in a separate silo
  • Platform engineers building internal LLM observability who don't want to write custom instrumentation for every provider SDK change
  • Organizations evaluating multiple LLM observability backends and wanting to avoid vendor lock-in at the instrumentation layer
  • Python-heavy stacks — this is the primary supported language, so if your LLM calls are in Python, integration friction is low

Who should NOT use this

  • Teams not on Node/TypeScript or other non-Python stacks — check current SDK parity before assuming full feature support outside Python
  • Anyone expecting a full observability platform out of the box — this is instrumentation, not dashboards, alerting, or evaluation tooling. You still need a backend and likely still need Traceloop's paid tier or a competitor for the analysis layer
  • Organizations with strict requirements around vendor-neutral governance (e.g., wanting a CNCF-graduated project) — maturity status here isn't publicly confirmed, so procurement teams requiring that box-check should verify directly
  • Teams needing enterprise support contracts today — if that's a hard requirement, get Traceloop's commercial terms in writing before committing engineering time

Alternatives

  • OpenInference (Arize) — similar OTel-based instrumentation approach, backed by Arize's observability platform, strong on evaluation-focused traces.
  • Langfuse — open-source LLM observability with its own tracing SDK (not OTel-native), stronger built-in UI/analytics if you don't already have an OTel pipeline.
  • Helicone — proxy-based LLM observability, simpler to bolt on if you don't want to touch application code at all, less flexible than SDK instrumentation.

Pricing

OpenLLMetry itself is fully open source under Apache-2.0 — free to use, self-host, and modify with no feature gating in the instrumentation layer. Cost enters if you route traces to Traceloop's hosted backend, which has its own commercial pricing (not covered here — check Traceloop's site directly). Sending spans to a self-hosted or third-party OTel-compatible backend costs you nothing beyond that backend's own pricing.

Frequently asked

What is OpenLLMetry?+
OpenLLMetry is an open-source observability layer for LLM applications, built as a set of OpenTelemetry instrumentations. It auto-instruments calls to LLM providers (OpenAI, Anthropic, Cohere, etc.
Who builds OpenLLMetry?+
Maintained by Traceloop, a company selling a commercial LLM observability/evaluation platform. OpenLLMetry is their open-core wedge — free instrumentation that funnels naturally into their paid backend, though the OTel-native design means you're not locked in if you skip it.
Is OpenLLMetry production ready?+
7,253 GitHub stars, active commit history through mid-2026, and a mature release number (0.62.1) point to a project past the "weekend hack" stage and into steady iteration — the 0.x versioning after 60+ releases is more about SemVer discipline than instability. Apache-2.
Who should use OpenLLMetry?+
Teams already running OpenTelemetry pipelines who want LLM spans to show up alongside existing service traces instead of in a separate silo Platform engineers building internal LLM observability who don't want to write custom instrumentation for every provider SDK change Organizations evaluating multiple LLM observabil
Who should not use OpenLLMetry?+
Teams not on Node/TypeScript or other non-Python stacks — check current SDK parity before assuming full feature support outside Python Anyone expecting a full observability platform out of the box — this is instrumentation, not dashboards, alerting, or evaluation tooling.
What are the alternatives to OpenLLMetry?+
OpenInference (Arize) — similar OTel-based instrumentation approach, backed by Arize's observability platform, strong on evaluation-focused traces. Langfuse — open-source LLM observability with its own tracing SDK (not OTel-native), stronger built-in UI/analytics if you don't already have an OTel pipeline.
How much does OpenLLMetry cost?+
OpenLLMetry itself is fully open source under Apache-2.0 — free to use, self-host, and modify with no feature gating in the instrumentation layer. Cost enters if you route traces to Traceloop's hosted backend, which has its own commercial pricing (not covered here — check Traceloop's site directly).