aacyn

eBPF observability without the instrumentation. 30 seconds to your first dashboard.

Hubble tells you what talked to what. aacyn tells you if it's breaking.

Quickstart

# 1. Clone the repo
git clone https://github.com/bneb/aacyn
cd aacyn

# 2. Install in any Kubernetes cluster
helm install aacyn ./charts/aacyn

# 3. Forward the dashboard
kubectl port-forward svc/aacyn-dashboard 3000:3000

# 4. Open http://localhost:3000

Requires kernel >= 5.15 and a Kubernetes cluster with privileged DaemonSet support.

Architecture

eBPF probes (kernel) → libaacyn.so (C, SoA columnar store) ←bun:ffi→ Elysia API (TS/Bun) → Next.js dashboard / Grafana plugin

Two primary languages, one data plane: C for the hot path (ingest, SIMD scans, eBPF consumer) and TypeScript for the control plane (API, archiver). An experimental Grafana data source plugin lives in grafana-plugin/.

How it compares

	aacyn	Hubble	Pixie	Coroot	Beyla
Golden Signals	✅ Built-in	❌ Manual	❌	✅	✅
Pre-configured alerts	✅ 5 rules	❌	❌	⚠️ SLO only	❌
SLO tracking	✅ Error budgets	❌	❌	✅	❌
Distributed tracing	✅ eBPF spans	❌	✅	✅	✅
gRPC visibility	✅ Kernel-level	✅	✅	✅	✅
K8s pod enrichment	✅	✅	✅	✅	✅
OTLP ingest + export	✅ Both	❌	❌	✅ Export	✅ Export
Forwarders	Datadog, Splunk, OTLP	❌	❌	❌	❌
No external DB	✅	✅	❌	❌ ClickHouse	✅
SIMD columnar store	✅ 5M events/s	❌	❌	❌	❌
Apache 2.0	✅	✅	✅	✅	✅

Full comparison →

Features

• Golden Signals — Latency, traffic, errors, saturation per-service out of the box. No dashboards to build.
• SLO Tracking — Define latency and error rate targets. Track error budgets and burn rates. Get alerted before you blow your SLO.
• Distributed Tracing — Automatic span generation from eBPF probes with W3C traceparent propagation. Waterfall view in the dashboard.
• HTTP & gRPC visibility — Kernel-level HTTP/1.1, HTTP/2, and gRPC parsing. Method, path, status code, service name — no sidecar needed.
• Kubernetes pod enrichment — Topology edges show pod names, namespaces, and deployments — not raw IPs.
• Pre-configured alerts — Ready-to-use alert rules for latency spikes, error bursts, and connection failures.
• Topology graph — Live, auto-discovered service dependency graph rendered in Canvas 2D. Updates as pods come and go.
• Grafana plugin — Native data source plugin. Drop aacyn metrics into existing dashboards.
• OTLP ingest + export — Accepts and forwards OpenTelemetry traces via HTTP/protobuf. Bridge existing instrumentation alongside eBPF data.
• Forwarders — Ship telemetry to Datadog, Splunk, or any OTLP-compatible collector.
• Zero instrumentation — No SDKs, no sidecars, no code changes. One Helm install, done.

Install

• Kubernetes — git clone this repo, then helm install aacyn ./charts/aacyn. Requires kernel >= 5.15, privileged DaemonSet.
• Docker demo — docker compose up in docker-demo/. Full stack on a single machine — no K8s required.

Tech Stack

Layer	Technology	Role
Agent	C + eBPF	Zero-footprint kernel-level telemetry capture
Storage	C + mmap	Columnar event store @ 5M events/sec/node
Query	C + SIMD	AVX-512/NEON scans (286us across 5M events)
API	Elysia + Bun	Type-safe control plane with Zod validation
UI	Next.js + Canvas 2D	Live topology graph with animated edges
Forwarding	Pluggable	Datadog, Splunk, OTLP

Monorepo Layout

aacyn/
├── native/          # C engine + eBPF probes (libaacyn)
├── ts/              # Control plane & UI (TypeScript / Bun)
│   ├── apps/
│   │   ├── web/     # Next.js dashboard
│   │   └── api/     # Elysia API server
│   └── packages/
│       ├── sdk/     # Shared types & telemetry schemas
│       └── ui/      # Shared component library
├── grafana-plugin/  # Grafana data source plugin
├── charts/          # Helm chart (Kubernetes)
├── docker-demo/     # Demo stack
└── deploy/          # systemd unit for bare metal

License

Apache 2.0 — see LICENSE for details. Current release: v1.0.0-dev.

Community

• GitHub Discussions — questions, ideas, show and tell
• Discord — real-time chat
• Contributing — how to build, test, and submit changes