HorusUse case
Investigate production incidents with evidence, not guesswork
Horus turns logs, metrics, queues, database state, and recent code changes into a deterministic, evidence-backed investigation — so you find what changed and why, faster.
Most incidents are still reconstructed by hand: tabs across dashboards, a grep through logs, a guess at the deploy that broke it. Horus turns those scattered signals into a single, evidence-backed investigation you can read top to bottom.
The problem: incidents are reconstructed by hand
- Signals live in different tools — logs, metrics, queues, database — with nothing linking them.
- The code that changed is in a fourth place, disconnected from the symptom.
- Every responder re-derives the same picture under pressure.
What a real investigation needs
- Evidence from runtime — logs, metrics, queue and database state.
- Source context — the symbols and files implicated by that evidence.
- A timeline — what changed, and whether it lines up with when symptoms began.
How Horus runs an investigation
Horus runs locally in your repo, builds a code graph, connects your runtime sources, then investigates an incident hint:
horus investigate "checkout latency spike"It correlates the symptom with recent changes, queue and database state, and the code path, then resolves to the most likely symbols and files.
What you get: an evidence-backed report
| Section | What it answers |
|---|---|
| Timeline | When the symptom started and what changed around it |
| Evidence | The logs, metrics, and state that support the finding |
| Suspected changes | Recent commits and symbols most likely involved |
| Affected owners | Whose code and which services are implicated |
| Next checks | Concrete steps to confirm or rule out the hypothesis |
Deterministic by design
The same inputs produce the same report. Re-render a saved investigation with horus replay, or draft a postmortem from it — no opaque, one-off LLM output.
Works with your stack
Horus reads runtime evidence through connectors (Elasticsearch, Grafana, Axiom, Sentry, MongoDB, BullMQ/Redis) and combines it with source intelligence. Follow the investigation walkthrough for a full example.