What it does

Superlog is a self-hosted observability workspace that ingests traces, logs, and metrics via OTLP, collapses the noise into incidents, and surfaces them in a React-based web UI. It runs on Postgres and ClickHouse, with background workers handling grouping and agent orchestration. There’s also a hosted cloud edition if you’d rather not babysit Docker.

The interesting bit

The “agentic” angle isn’t just marketing garnish. Superlog includes pluggable “agent runner interfaces” and a default community runner that records local incident summaries. The pitch is that these agents investigate and potentially self-heal problems without a human waking up. Whether they actually succeed at that is left as an exercise to the operator — the framework is there, the intelligence is presumably yours to configure.

Key highlights

• Full OpenTelemetry ingest via OTLP proxy; no proprietary SDK required
• ClickHouse-backed queries for telemetry data, Postgres for metadata and schema
• Modular monorepo: separate web, API, proxy, and worker apps
• “Skills” installable via npx skills add for coding-agent integration
• Apache 2.0 licensed, with Y Combinator backing (P26 batch)

Caveats

• The README mentions AI agent self-healing but the open-source edition only ships a community runner that “records a local incident summary” — the fancy autonomous remediation appears to be cloud-edition territory or roll-your-own
• Requires Node 20+, pnpm 9+, and Docker; not a lightweight drop-in
• 667 stars suggests early traction, but the project is young enough that rough edges are likely

Verdict

Worth a look if you’re already on OpenTelemetry and want an open-core alternative to Datadog/New Relic with room to grow into agentic workflows. Skip it if you need battle-tested, fully autonomous remediation out of the box — this is a framework with promise, not a finished oracle.