HorusComparison
Horus vs observability tools: dashboards show signals, Horus builds investigations
Observability tools show metrics, logs, traces, and dashboards. Horus connects those signals with code changes, queues, ownership, timelines, and evidence for AI-assisted incident investigation.
Observability tools — Grafana, Elasticsearch, Datadog, Prometheus — are excellent at collecting and displaying signals. Horus doesn't replace them. It reads from them and turns those signals into an investigation.
Observability tools are the source of truth for signals
Metrics, logs, and traces belong in your observability stack. Horus assumes they're there and treats them as inputs — it doesn't store metrics or try to be your dashboard.
The missing layer: turning signals into investigation context
A dashboard shows a latency spike. It doesn't tell you which deploy preceded it, which queue is backed up, which code path is implicated, or who owns it. That correlation is the work — and it's what Horus does.
Why dashboards alone are hard for AI coding agents
- Dashboards are visual; agents need structured, linked data.
- Signals are siloed per tool; an agent must query each and reconcile formats.
- No code linkage: a metric doesn't point at the function or the commit that changed it.
What Horus adds
- Evidence-backed reports instead of a wall of panels.
- Source intelligence linking signals to the code they implicate.
- Timeline of changes correlated with symptoms.
- Owner mapping for the affected symbols and services.
- Suggested next checks an agent or engineer can act on.
Not an APM replacement
Horus doesn't store metrics, replace your dashboards, or page on-call. Keep your observability stack — Horus consumes it to build investigations.
How Horus works with Elasticsearch, Grafana, Axiom, MongoDB, PostgreSQL, BullMQ/Redis, Sentry, and source code
Horus reads runtime evidence through connectors and combines it with a local code graph:
| Source | What Horus reads |
|---|---|
| Elasticsearch | Logs and search-indexed events |
| Grafana | Metrics and dashboard panels |
| Axiom | Logs and event streams |
| MongoDB | Database state relevant to the incident |
| PostgreSQL | Application and operational state from allowlisted tables |
| BullMQ / Redis | Queue topology, job state, and cache/lock state |
| Sentry | Grouped exceptions seeded to the implicated code |
| Source code | Symbols, structure, and recent changes via source intelligence |
See connectors and source intelligence for how each source is wired in.
The one-line difference
Observability tells you what is happening. Horus helps explain what changed, why it may be happening, and what to check next.