OpenSearch Observability Stack
OpenSearch Observability Stack is an open-source, OpenTelemetry-native observability platform (GitHub) for monitoring services, infrastructure, and AI agents. Install locally via Docker Compose - traces, logs, Prometheus metrics, service maps, and agent tracing out of the box.
- Collect via OTLP
- Log-trace correlation
- Search with PPL and PromQL
- Auto-generated service maps
- RED metrics from traces
- Latency and error tracking
- Trace LLM calls, token usage, and tool use
- Agent execution graphs
- Python and JavaScript SDKs
- Custom dashboards
- Anomaly detection
- Slack, email, webhook alerts
See it in action
Section titled “See it in action” Live Playground Try the Observability Stack live - no installation required.
Quickstarts
Section titled “Quickstarts” Install & Explore Install the stack and explore traces already flowing from the built-in example services.
Send Your First Traces Instrument an app with OpenTelemetry and see traces in dashboards.
Trace an AI Agent Instrument an AI agent with GenAI semantic conventions and visualize the execution graph.
Query language: PPL
Section titled “Query language: PPL”The Observability Stack is powered by Piped Processing Language (PPL) - a pipe-based, human-readable query language purpose-built for observability. PPL is the native language for querying logs and traces, giving you a single, consistent syntax across signal types.
source = logs-otel-v1*| where severityNumber >= 17| stats count() as errors by `resource.attributes.service.name`| sort - errorsWith 50+ commands and 200+ built-in functions, PPL covers everything from simple filtering to machine learning anomaly detection - all in a pipeline you can read top to bottom.
Learn PPL Overview, command reference, function reference, and real-world observability examples with live playground links.
Why Observability Stack?
Section titled “Why Observability Stack?”- Open source: Fully open source, no vendor lock-in, self-host everything
- OpenTelemetry-native: All data ingestion uses OTel protocols and semantic conventions
- PPL-native: Pipe-based query language across logs and traces - no query DSL, no steep learning curve
- GenAI-first: Purpose-built views for AI agent tracing using standard
gen_ai.*attributes - Local-first: Runs entirely on your machine via Docker Compose - no cloud account required
- Production path: Same components (OpenSearch, Prometheus, OTel Collector) scale to production
Community
Section titled “Community”- GitHub - issues, PRs, and discussions
- Contributing guide - how to contribute