Debug production issues by tracing symptoms to code

Debugging production usually means jumping between a dashboard, a log viewer, the queue UI, and the repo — holding the connections in your head. Horus makes those connections for you, tracing a symptom back to the code that caused it.

The problem: debugging prod means tool-hopping

The signal that something is wrong and the change that caused it almost never live in the same place. You reconstruct the link manually, every time, and the context evaporates when the incident closes.

Common symptoms Horus helps with

• Slow endpoints and latency spikes (e.g. a slow checkout).
• Failed or stuck background jobs and queue backups.
• Regressions that started after a specific deploy.
• Database-state issues surfacing as user-facing errors.

How Horus connects symptom → evidence → code

• Pulls runtime evidence for the symptom from your connected sources.
• Maps it onto a code graph and recent changes via source intelligence.
• Builds a timeline so the symptom is anchored to what changed.
• Resolves to the symbols and files most likely responsible.

Example: a failed background job

horus investigate "payment webhook jobs failing"

Horus correlates the job failures with queue state, recent commits to the handler, and related log lines — then points at the change and the owner instead of leaving you to grep.

What you get

A deterministic, evidence-backed report: timeline, supporting evidence, suspected changes, affected owners, and concrete next checks. Re-render it any time with horus replay, or hand it to an AI coding agent as grounded context.

See how Horus reads each source in connectors, how it understands code in source intelligence, or walk a full example in the investigation walkthrough.