1. What each tool actually watches
| Tool | Watches | Misses |
|---|---|---|
| Sherlocks.ai | Metrics, logs, traces, change events, tickets, cloud services, databases, queues and Kubernetes | – |
| k8sgpt | Kubernetes objects & events | Everything outside the cluster |
| RunWhen | K8s manifests and upgrade state | App-level signals and multi-cloud context |
The short version: k8sgpt and RunWhen know a lot about what’s inside your cluster. Sherlocks looks at the whole system—cloud services, queues, the noisy neighbour DB, you name it.
2. What happens when something breaks
Sherlocks.ai
- Builds a timeline from every signal it can find
- Offers a root-cause hypothesis in plain English
- Suggests (or auto-executes) a fix and leaves an audit trail
k8sgpt
- Explains what’s wrong with a pod or deployment
- Spits out one-off commands to try next
- Stops at the edge of the cluster
RunWhen
- Runs best-practice rules and upgrade playbooks
- Great for hygiene, less so for surprise outages
3. How you interact with each tool
Sherlocks drops into your Slack or Zoom bridge, answers follow-ups, tracks action items and drafts the RCA before the smoke clears.
k8sgpt and RunWhen live in a CLI or their own dashboard-you run them, read the output, and take it from there.
4. Proactive vs. reactive
Sherlocks never sleeps. It scans telemetry around the clock and raises a flag before the pager barks.
k8sgpt and RunWhen kick in after you ask them to look or on a scheduled scan.
5. Plug-and-play reality check
| Feature | Sherlocks.ai | k8sgpt | RunWhen |
|---|---|---|---|
| Datadog / CloudWatch / Grafana integration | ✔️ | ❌ | ❌ |
| Hybrid / on-prem deploy | ✔️ | ✔️ (binary) | ⭕ |
| SOC 2 roadmap | In progress | Community OSS | In progress |
| Human-in-loop approvals | ✔️ | N/A | Rule-based |
The takeaway
- k8sgpt is a sharp cluster doctor-excellent when your YAML goes sideways.
- RunWhen automates the routine health checks and upgrades we all forget.
- Sherlocks.ai plays the role of full-stack SRE teammate. It sees the whole system, reasons across services, and helps drive the incident to closure-whether that means a kubectl rollback, a database failover or a quick note in PagerDuty.
If that sounds useful, let’s chat. I’m always happy to walk through a live demo on your own stack.
Stay stable, ship faster. 🚀