Skip to main content
MoleSignal is a self-hosted, OpenTelemetry-native observability platform. Logs, metrics, and traces share one storage layer (Parquet on object storage), one query engine (DataFusion + Arrow), and one metadata layer (Postgres) — so a trace, its logs, and the host metric of the same minute are actually connected, not stitched together by hand.
MoleSignal is pre-1.0 and under active development. Expect rough edges, and please open an issue — every report shapes the next release.

​
Why MoleSignal

Today’s telemetry tools force a bad trade-off:
  • Commercial SaaS (Datadog, New Relic, Splunk) correlates the three signals, but the bill grows linearly with traffic — a mid-size team easily pays US$2k–10k/month for 100 GB/day.
  • Open-source stacks (Loki + Mimir + Tempo, or ELK + Prometheus + Jaeger) are free, but logs, metrics, and traces live in three separate stores with three query languages. The “trace ↔ log ↔ host metric” jump everyone needs during an incident has to be stitched by hand.
MoleSignal takes the third path: the three signals are correlated at the data plane, not at the dashboard plane. Self-hosted, so your bill is the S3 cost.
Commercial SaaSOSS stackMoleSignal
100 GB/day cost~US$2k+/moinfra onlyinfra only
Three signals, same storage✅ (their cloud)❌ 3 stores✅ Parquet + DataFusion
Cross-signal correlation✅ (paid)⚠️ manual copy-paste✅ native
Data ownershiptheir cloudself-hostedself-hosted
Setup time5 min6 hours+1 cmd docker compose up
Real-time alerts (<1s)yesnoyes (kind: realtime)

​
Get started

Quickstart

Spin up a sandbox, send your first telemetry, and run a cross-signal query in minutes.

Core concepts

Orgs, streams, signals, and how correlation works under the hood.

Ingesting data

10 drop-in ingest protocols: OTLP, Prometheus, Loki, Elasticsearch, and more.

Querying

Full SQL and a PromQL subset over one engine, across all three signals.

​
Key capabilities

  • Cross-signal correlation — drill metric → trace → log → host and back without losing context.
  • 10 ingest protocols — drop-in replacements for OTLP, Prometheus remote_write, Loki, Elasticsearch _bulk, Syslog, and more.
  • One query engine — DataFusion SQL with joins/CTEs/window functions, plus a PromQL subset.
  • Alerting — scheduled, realtime (fires <1s), and anomaly rules with escalation policies.
  • Multi-tenant & secure — planner-level org isolation, API tokens, audit log, field-level encryption.
  • Operations — 6 stateless roles, single binary, Kubernetes manifests, Prometheus /metrics.