AI SRE: Find. Fix. Ship.
Observability and incident intelligence for AI agents running in production — built on OpenTelemetry and SigNoz.
AI agents are moving into production faster than the tooling to watch them.
Traditional APM tells you a request returned 200 OK. It doesn't tell you that the agent hallucinated an answer, looped on the same tool call five times, picked the wrong tool entirely, or silently burned through its token budget before failing halfway through a task.
By the time a human notices, the incident has already happened.
VOID instruments AI agent executions with OpenTelemetry and turns that telemetry into incident intelligence.
Application → VOID SDK → OpenTelemetry → OTLP → SigNoz → Incident Analysis → Engineering Report
Today, that means a lightweight SDK that emits standards-compliant traces, and an analyzer that turns those traces into engineering-grade incident reports — repeated tool calls, failed executions, excessive token usage, hallucinations, and wrong-tool selections, all flagged with evidence and a recommendation.
Tomorrow, the same telemetry pipeline powers autonomous incident response — filing GitHub issues, notifying Slack, triggering PagerDuty, opening Linear tickets, and coordinating repair agents — without changing a single line of the instrumented application.
The SDK's only job is to emit telemetry. Detection, analysis, and remediation live entirely downstream, so today's integration keeps working as the intelligence layer above it gets smarter.
| Repo | Description |
|---|---|
void-demo |
Interactive product demo and pitch deck — see the SDK, telemetry pipeline, and incident analyzer in action |
VOID started as a hackathon project and is being built out as a real product. Issues, feedback, and contributions are welcome on the repositories above.
