About Mintlify
Mintlify is building the intelligent knowledge platform for the AI era. Their platform transforms documentation from static content into living, AI-native systems of knowledge. By integrating intelligence across the documentation lifecycle—from authoring and maintenance to discovery and user interaction—Mintlify enables teams to create self-updating docs, AI-powered assistants, and context-aware knowledge experiences.
Trusted by companies like Anthropic, Microsoft, and Coinbase, Mintlify powers documentation for millions of developers each month and supports how modern teams build and share knowledge.
The Challenge
Mintlify operates at a high development velocity. As a product-led startup serving developer teams, engineers ship frequently and infrastructure changes are pushed regularly.
With this pace of development, production alerts are a regular part of the engineering workflow. Mintlify receives around 100 engineering alerts per day across its systems. When alerts fire, the on-call engineer must investigate the alert manually by opening it in Datadog, searching logs and metrics, identifying impacted services, reviewing recent deploys or pull requests, and determining whether action is required.
Even when alerts do not require action, the investigation process still requires time and attention. With alerts arriving continuously throughout the day, the on-call engineer repeatedly gathers context across multiple tools to understand what is happening in the system.
Manual investigation typically takes around five minutes per alert, which creates a substantial operational workload over the course of a day.
Mintlify wanted to move toward a reliability model where engineers could immediately understand alerts with full context and reduce the manual effort required during on-call.
The Solution
Mintlify implemented Deeptrace within their existing engineering workflow. Deeptrace integrates with Datadog, GitHub, and Slack and automatically investigates alerts as they occur.
When an alert appears in Slack, Deeptrace posts an investigation memo that includes relevant logs and metrics from Datadog, related deploys and pull requests, affected services, and an analysis of potential causes.
Instead of opening multiple tools to begin debugging, engineers review the Deeptrace memo directly in Slack to determine whether further investigation or action is required.
Engineers can also interact with Deeptrace in Slack to request additional analysis or explore related signals across their observability data and codebase.
After integrating Datadog, GitHub, and Slack, Deeptrace immediately began investigating alerts and supporting engineers on-call in under an hour.
The Result
Deeptrace is now part of Mintlify's on-call workflow.
With alerts investigated automatically and context delivered in Slack, engineers can review the Deeptrace memo to determine whether an alert requires action instead of manually investigating each alert in Datadog.
Previously, investigating alerts took around five minutes per alert. With Deeptrace acting as the first responder, engineers estimate their time to resolution is now closer to one minute, as they rely on Deeptrace's output directly in Slack—saving roughly eight hours of on-call work per day.
Engineers can review alerts with the relevant investigation already attached, reducing the need to manually gather information across multiple systems and lowering the operational overhead associated with on-call debugging.
| Metric | Impact |
|---|---|
| Daily alerts | ~100 across systems |
| Investigation time per alert | 5 min → 1 min |
| On-call time saved | ~8 hours per day |
| Time to deploy | Under 1 hour |
“We get around 100 alerts a day. If we manually debugged each one in Datadog it would take about five minutes — with Deeptrace we just have to read a message. That's saved us roughly eight hours of on-call time every day.”
Nick Khami
Engineering Manager, Mintlify