· APACHE SOFTWARE FOUNDATION · OPEN SOURCE · COMMUNITY-DRIVEN

Observability for cloud-native and AI.
And AI for observability.

Apache SkyWalking puts distributed tracing, metrics, logs, profiling and alarms in one platform — across the services you run and the GenAI models they call, down to tokens and cost. 10+ language agents and eBPF on Kubernetes. Ask it in plain language, or investigate from the AI client you already use.

Username: skywalking Password: skywalking Go to UI Preview on Grafana
Star on GitHub
demo.skywalking.apache.org SkyWalking services dashboard with topology and active alarms
SkyWalking trace explorer with span waterfall
Distributed tracingSpan-by-span waterfalls.
SkyWalking logs explorer
LogsTrace-correlated log explorer.
SkyWalking metrics and dashboards
Metrics & dashboardsCustomizable widgets and MQE.
SkyWalking RBAC and alerting
RBAC & alertingRole-based access and alarm rules.
horizon · AI Assistant
Bring your own LLM

No frontier model required. Vendor-neutral — any OpenAI-compatible endpoint or Amazon Bedrock — and runs efficiently on a cost-effective model like DeepSeek or a mid-tier Claude Sonnet. The observability know-how lives in the tools, metric catalog and playbooks; the model orchestrates and narrates. Off by default.

NEW · AI ASSISTANT

Ask questions in natural language

Horizon's built-in assistant answers questions about your running system — "what's unhealthy right now?", "investigate latency for checkout" — with no MQE, PromQL or TraceQL to write. It replies with an ordered narrative built from the same charts, topology and tables as the rest of the UI, not a wall of text.

  • Answers you can see. Inline charts, top-N lists, tables and a focused one-hop topology — drawn by the same widgets as the dashboards.
  • Guided root-cause. Playbooks walk root service → calling chain → error stack, read live Kubernetes pod logs, and can propose a profiling task only you approve.
  • Read-only, your permissions. Every action re-checks a read verb you already hold — it never sees more than you can, and never edits anything.
AI Assistant docs Meet Horizon UI
codex · horizon
The pattern

An entrance, not a scraper. The client calls an authorised capability and reasons from what comes back — it never clicks through the pages. SkyWalking keeps the semantics, permissions and presentation; the client keeps the model. From AI Clients as First-Class Interfaces.

NEW · YOUR OWN AI CLIENT

Investigate from the AI client you already use

SkyWalking exposes the same investigation capabilities over MCP, so an agent such as Codex can query your running system directly. In the ChatGPT desktop app, its own topology, deployment and chart components mount inside the conversation — interactive, not screenshots. In the CLI, the same capture comes back as structured evidence for a terminal workflow.

  • One implementation, not two. The built-in assistant and an outside agent run the same investigation capabilities against the same live data — there is no second definition of a service, a metric or a topology to keep in sync.
  • Your permissions, still enforced. Connecting takes a permission of its own, and every capability then re-checks that you already have read access to the matching screen. Nothing on this path can change anything — profiling can be proposed, never started.
  • The model stays on your side. Your agent brings its own model and its own subscription; SkyWalking only issues a read-scoped token, and never makes an LLM request itself.
Connect your agent
Interactive · sample data Apache SkyWalking topology rendered as an interactive 3D scene you can orbit, zoom and click
SEE IT IN 3D

Explore your topology in 3D

Walk your services and cloud-native infrastructure as a living 3D scene — orbit, zoom and click any node to inspect it. It opens as a full page and runs entirely in your browser from a frozen demo snapshot: no server, no login, no setup.

Open the 3D map full page
WHAT YOU GET

One platform for traces, metrics, logs, profiles and GenAI

An observability platform designed for cloud-native systems. Auto-instrument the languages you run, observe the kernel via eBPF, and store the volume natively in BanyanDB.

Tracing

End-to-end distributed tracing

Native plus Zipkin v1/v2 formats. Auto-instrument 10+ languages, eBPF on Kubernetes, and dependency analysis across services and APIs.

Learn more →
Metrics

Metrics pipeline at any scale

OTLP, native, OpenCensus, Telegraf and Zabbix flow through one MAL script pipeline — 100B+ data points per cluster.

Learn more →
Logging

Logs with trace correlation

Collect from disk or network, auto-bind trace context, extract metrics, sample by policy, and forward anywhere.

Learn more →
Profiling

Profiling, in-agent and eBPF

CPU and async profiling bundled in language agents. eBPF profiler on Rover for Kubernetes deployments.

Learn more →
GenAI

LLM and GenAI monitoring

Latency, traffic, success rate, input/output tokens and estimated cost, per provider and model — from SkyWalking, OTLP or Zipkin traces.

Learn more →
And the rest

See all capabilities

Every server, agent, database and integration in the project — each with documentation for every released version.

Browse documentation →
ECOSYSTEM

Built for the stack you already run

Language agents, service mesh, eBPF, a native observability database, and broad telemetry-format support — vendor-neutral and Apache-governed.

Agents for your stack

Java, Go, Python, Node.js, PHP, Rust, Ruby, .NET, LUA, C++ and Client JavaScript — actively maintained.

11 language agents →

eBPF with Rover

Monitor and profile Kubernetes deployments and diagnose CPU and network performance with eBPF.

Rover docs →

BanyanDB

A native observability database built to ingest, analyze and store telemetry at scale.

BanyanDB docs →

AI-assisted troubleshooting

Ask Horizon UI to investigate in natural language — LLM-driven root cause, read-only and grounded in live data — plus ML baselines and anomaly detection.

AI on the blog →
Telemetry supported natively OpenTelemetry (OTLP) Zipkin v1/v2 Prometheus OpenCensus Telegraf Zabbix Fluentd / Fluent Bit Apache Kafka
LATEST

Events & blogs

GET STARTED

Ready to get started?

Spin up the Showcase to see Apache SkyWalking in action, or install the platform — then join a community built by 1017 contributors.

Quick start All releases

Questions or bugs?

Open a discussion or file an issue on GitHub — feature requests, questions and bug reports all welcome.

GitHub Discussions →

Join our Slack

Chat with maintainers and users in real time. ASF self-signup only works for @apache.org addresses, so ask the dev mailing list and a maintainer will invite you.

dev@skywalking.apache.org →

Follow on Twitter

Announcements, releases and the latest features — follow @ASFSkyWalking for project news.

@ASFSkyWalking →