We built Devtract because we kept getting paged at 3am for problems that took an hour to diagnose and five minutes to fix.

Founder Story

Before Devtract, Marcus Chen spent five years in platform engineering at a logistics platform running 200-plus microservices with a deployment cadence of 40 to 60 releases per week. MTTR was consistently above 45 minutes, not because the team lacked skill, but because every incident started the same way: an alert fires at 3am, someone opens a Slack war room, and five engineers begin independently grepping logs, each trying to form the same hypothesis.

The problem was not the engineers. The signal was scattered across Datadog, GitHub, and Prometheus, and correlating it manually consumed the first 30 to 40 minutes of every incident. By the time someone found the root cause, the actual fix took five minutes. Deploy a bad config, service starts OOMKilling on startup, roll back the config. Two minutes of remediation. Forty minutes of detective work.

In late 2023, Marcus built an internal correlation tool that automated the diagnosis step. It read from existing telemetry APIs, opened a time window around each alert, and ranked overlapping anomaly clusters by statistical evidence. MTTR dropped by more than half within the first month of internal use. In early 2024, he left to turn that internal tool into Devtract, incorporated in Austin, TX.

Devtract is not an APM. It does not replace Datadog or Prometheus. It is the intelligence layer that reads from those tools and does the correlation work an engineer would otherwise do manually at 3am. We are a small team and every feature decision starts with the same question: would this help someone who just got paged?