A refreshed AI Assistant, AWS DevOps Agent for MCP, and service names for S3 logs

Improvements Fixes

TL;DR

A denser AI Assistant chat, MCP support for the AWS DevOps Agent, and automatic service names for S3 logs — plus Kubernetes, dashboard, and RUM fixes.

A refreshed AI Assistant, AWS DevOps Agent for MCP, and service names for S3 logs

Improvements

  • AI Assistant: Refreshed chat experience — a denser layout, syntax-highlighted tool inputs and outputs, and a compact tool-call timeline that stays expanded while a response is running and auto-collapses once it finishes. Switching between chats is smoother, with no flicker on fast switches.
  • MCP
    • The AWS DevOps Agent can now connect to your hosted MCP server, with registration handled as a standard OAuth PKCE client
    • Questions about structured fields like HTTP status codes now filter on the attribute itself instead of doing a plain-text search of the log body, so you get the right results instead of empty or misleading ones
  • Discover
    • The Hosts Processes tab now shows friendly process names for Java services — the jar name instead of the full JVM command line, so you can tell your services apart
    • Charts on the Kubernetes Overview can now open in fullscreen for incident response and deep dives
  • Logs & Traces: Duration fields can be displayed in milliseconds or seconds in queries instead of raw nanoseconds, so you no longer have to count zeros to read a value
  • Alerting
    • MS Teams notifications now work with Power Automate workflow webhook URLs and render as Adaptive Cards, replacing the retired O365 connectors
    • New API support for creating and attaching notification channels programmatically
  • Integrations: Logs ingested from Amazon S3 now always get a service name: an explicit per-bucket rule wins, otherwise built-in parsing pulls the web ACL name from AWS WAF logs and the load balancer name from ALB access logs, falling back to the bucket name. So WAF and ALB logs land under meaningful service names out of the box.

Fixes

  • Dashboards: Sorting a Table panel column descending placed rows with missing values ("-") at the top instead of the bottom
  • Discover
    • The Kubernetes page was slow to reflect pod state changes — newly running pods and CrashLoopBackOff states now appear promptly
    • The Applications (RUM) Errors by Type, City, and Country views showed "Unknown" errors with no associated trace, polluting error analytics