This Month in AWS Observability: July 2026

Introduction

July was a busy month for AWS Observability. We launched features to make the telemetry you already collect more actionable, and take the operational work of collecting it off your hands. Log analytics moved closer to action with alarms that run straight from a log query and enrichment that happens at ingestion. Application-level observability started capturing errors and deployment events on its own. Visibility into AI coding agents arrived as a new observability category. Managed collection removed the agents you used to run for scraping Prometheus metrics. Underneath all of it, OpenTelemetry and Prometheus continue to serve as first-class foundations.

Here’s what’s new across Amazon CloudWatch, Amazon Managed Service for Prometheus, Amazon Managed Grafana, and AWS DevOps Agent.

To see live demos of these launches, register for the next “I didn’t know Amazon CloudWatch could do that!” webinar on September 16. Missed earlier recaps? Catch up on This Month in Observability Jan-May 2026 and June 2026 blogs.

Log analytics: from query to action

The clearest theme this month is closing the distance between a log query and the action you take on it.

You can now create alarms directly from CloudWatch Logs queries. Write a query, set a threshold, and receive an alert with the surrounding log context when errors spike, all in a single workflow. There is no longer an intermediate step of building a metric filter or a custom metric first. Log alarms support the standard alarm actions you already use, including Amazon Simple Notification Service and Amazon EventBridge, so you can route them into the same notification and automation paths as any other CloudWatch alarm.

The new lookup processor for log enrichment adds context to log events at ingestion time by matching fields against a CSV lookup table you upload. You can map IP addresses to owning teams, error codes to readable descriptions, or user IDs to account metadata without writing custom enrichment logic. Because the enrichment happens as data arrives, your queries, dashboards, and alarms all benefit from the added fields immediately.

CloudWatch Logs now supports Application Load Balancer logs as vended logs. Access, connection, and health check logs from your Application Load Balancers stream directly to CloudWatch Logs, where you can query them with Logs Insights, build metric filters, and watch them in real time with Live Tail. Telemetry enablement rules auto-configure logging for existing and new load balancers across your organization, so you do not wire up each one by hand.

CloudWatch Logs Insights also added 25 new query commands and functions in July, including statistical functions, sessionization, outlier detection, and null handling. Together, they let you perform more analysis directly in the query language instead of post-processing results elsewhere.

Cost-aware log storage

More logging in one place raises the question of storage cost, and July addressed that too. CloudWatch Logs Intelligent Tiering classifies log data into three tiers, Standard, Infrequent Access, and Archive Instant Access, based on how recently the data has been accessed. Data that has not been accessed for 30 days moves to Infrequent Access, and data untouched for 90 days moves to Archive Instant Access. When you query older data, it promotes back to Standard automatically. The query experience is the same across all three tiers, so you lower storage cost without changing how your team works. You turn it on at the account level.

Application observability

Application Signals now automatically captures errors, performance anomalies, and deployment events from your instrumented services, without additional code changes. It records exception and latency event snapshots, function-level performance data, and deployment events, so when something changes you can go directly to the Errors view and see whether a recent deployment introduced new exceptions. This is supported for Java, Python, and JavaScript services instrumented with AWS Distro for OpenTelemetry SDKs or the CloudWatch Observability add-on for Amazon Elastic Kubernetes Service (Amazon EKS). Events are delivered as logs and function metrics as OpenTelemetry metrics, so they sit alongside the rest of your observability data.

AI coding agent observability

Coding Agent Insights gives engineering leaders visibility into how AI coding tools drive value across an organization. It integrates with Claude Code, Codex, and GitHub Copilot, and collects OpenTelemetry metrics without extra instrumentation. The data is presented alongside your existing CloudWatch operational views. You can track spend trends, set token billing alerts, correlate agent adoption with commit throughput and pull request velocity, and identify which models deliver the best cost-to-output ratio. As teams adopt coding agents at scale, this gives them a way to measure return rather than guess at it.

OpenTelemetry and managed collection

Managed Prometheus collectors provide fully managed collection of Prometheus metrics with no agents to deploy or maintain. You provide a scrape configuration, and CloudWatch handles provisioning, scaling, and collection for you; enabling monitoring for Amazon EKS, Amazon EC2, Amazon ECS, Amazon MSK, and Amazon OpenSearch Service workloads without deploying or managing any agents. Metrics are delivered in OpenTelemetry format and are queryable with PromQL alongside your AWS vended metrics.

Related Blog Posts from AWS Cloud Operations Blog (July 2026)

The following posts were published on the AWS Cloud Operations Blog in July 2026:

Getting per-resource alarm notifications with Amazon CloudWatch – Himanshu Dewan, Ashish Kumar, Radheshyam Baliga Bantwal

Autonomous Root Cause Analysis for AWS Systems Manager Patch Failures Using AWS DevOps Agent – Rizwan Mohammed, Samir Behara, Shanmukha Jaya Harsha Alluri, Vinod Kisanagaram

Build bespoke operational workflows with AWS DevOps Agent custom SRE agents – Harish Mandhadi, Brent Everman, Joe Alioto, Janardhan Molumuri

Automate CI/CD troubleshooting with AWS DevOps Agent and GitHub – Purushotham G K, Abhishek Taparia

Extend Amazon CloudWatch Beyond Native Connectors with Cribl Stream – Gabriel Costa, Kishore Vinjam

How Amazon Achieved Full Stack Observability Across 400 Offices with Amazon OpenSearch Serverless – Shivansh Singh, Arijit Chakravorty, Bishr Tabbaa, Leonardo Quintero

Deploy OpenTelemetry Gateway on AWS: Monitoring Your Observability Pipeline – Ankita Saxena, Jyothi Madanlal

Turn Your Amazon CloudWatch Alarms into Actionable Signals – Helen Ashton, Gagandeep Singh

Using Amazon S3 Server Access Logs with Amazon CloudWatch Logs – Isaiah Salinas, Erik Weber

Conclusion

With July’s launches, log analytics closes the gap between query and action, storage tiering helps keep data affordable to keep, application observability captures the events that matter without extra code, AI coding agents become measurable, and managed collection takes the agent overhead off your plate, all built on OpenTelemetry and Prometheus foundations.

To get started with these capabilities:

• Create alarms directly from log queries in the CloudWatch console.
• Upload lookup tables to enrich logs at ingestion with CloudWatch Pipelines.
• Turn on Application Signals service events for automatic error and deployment capture.
• Configure Claude apps gateway telemetry to populate Coding Agent Insights.
• Replace self-managed collectors with managed Prometheus collectors.
• Stream Application Load Balancer logs to CloudWatch Logs with telemetry enablement rules.
• Turn on Intelligent Tiering for your log accounts to lower storage cost automatically.

For the full list of recent launches, visit the AWS What’s New page filtered to Amazon CloudWatch.

Looking for more?

Join our “I Didn’t Know Amazon CloudWatch Could Do That” webinar on September 16 to see these new features in action and use them to accelerate troubleshooting.

Authors