Amazon Quick builds an RCA brief for incidents from a single prompt using New Relic and Asana

AWS has published a step-by-step tutorial for building an AI incident triage agent on Amazon Quick, a service for creating agents with native integrations across the AWS ecosystem.

The problem the agent solves

When something breaks in production, the on-call engineer is under pressure: they need to find the cause quickly, gather context from several tools, write up an analysis and hand the task to a team. This process usually takes anywhere from 20 minutes to several hours, and the service is unavailable or degraded the whole time.

AWS proposes delegating the routine part of the investigation to an AI agent. The solution is built on three products:

• Amazon Quick: the agent orchestrator, which takes the prompt and manages the chain of tool calls
• New Relic MCP Server: the source of monitoring data, including metrics, traces, logs and alerts in real time
• Asana: the task tracker used to record the RCA brief and hand off to the engineering team

How the agent investigates an incident

The user describes the problem in a single free-form prompt. The Amazon Quick agent starts the investigation: it knows which tools to call and does so on its own.

First, the agent connects to the New Relic MCP Server. It requests data for the relevant time range, looks for anomalies in metrics and correlates traces with alerts. The MCP protocol lets the agent talk to the monitoring tool the same way a human engineer would, but without switching tabs or copying data by hand.

Once it has enough signals, the agent writes the RCA brief: an incident description, a timeline of events, the probable root cause, and direct links to specific logs and metrics as evidence. As its last step, the agent creates a task in Asana through the native integration, with the context already filled in and no manual entry.

"From a single prompt, the

Amazon Quick agent investigates an incident, assembles an RCA brief with links to evidence and creates a trackable Asana task ready for handoff," the AWS guide says.

Why the approach matters

A few details set this tutorial apart from similar ones:

• MCP as an enterprise standard: New Relic became one of the first major monitoring platforms with a native MCP Server. That signals to the market that the protocol is moving beyond chatbot demos.
• Native integrations without code: Amazon Quick connects the agent to Asana and New Relic through built-in connectors. No homegrown webhook wrappers.
• Audit trail by default: an RCA brief with direct links to evidence is not just a convenience but also documentation for postmortems and compliance audits.
• One prompt instead of 40 minutes: automating the first hour of response frees the engineer from routine data gathering.

Agents like this are especially in demand on teams where on-call engineers burn out on repetitive investigations. The standard pattern of "check the metrics → find the anomaly → form a hypothesis → create a ticket" can be fully automated, leaving the engineer only the decision.

What this means

The MCP protocol is moving into production tools: New Relic, Asana and Amazon Quick are no longer an experimental combination but a working pattern. For teams that already use these products, adding an AI agent to the on-call process becomes a matter of configuration rather than development from scratch.