Causely Feature Demo: Faster Investigations with AWS DevOps Agent
Anson McCook
October 7, 2026
When an incident fires, AWS DevOps Agent has to work out how your environment behaves before it can name a cause. It queries metrics, logs and traces one call at a time. That takes minutes, burns tokens, and can lead the agent to treat a correlated symptom as the cause. The delay is worst when an external service is at fault, because the agent has to rule out each of your own services first.
Causely runs inside your environment and keeps a live causal model of how failures spread across your topology. Your raw telemetry stays put. AWS DevOps Agent can connect to external MCP servers registered to its Agent Space and use the tools you allowlist from them during investigations. With the Causely MCP server connected, the agent gets that causal context in a single call.
We ran AWS DevOps Agent on two incidents, once without Causely and once with it:
| Incident | Without Causely | With Causely |
|---|---|---|
| Fault in internal service | ~10 min, 36 tool calls | ~5 min, 10 tool calls |
| Fault in external service | ~15 min | ~3 min |
When the fault was in an external service, Causely let the agent confirm the services were healthy and name the external service as the cause, without checking each service one by one.
Watch the video to see AWS DevOps Agent work both incidents with and without Causely. For results across more agents and scenarios, see the Causely benchmark.
