Datadog
Agent 7.79.0 shipped a macOS overhaul and a batch of new capabilities across networking, OTel, and observability.
macOS installation went system-wide — the Agent now installs as a LaunchDaemon under a dedicated _dd-agent user instead of a per-user LaunchAgent. DNS monitoring via libpcap, Cloud Network Monitoring via BPF filters, and CPU cache/platform metadata collection all landed alongside it. Per-user installs are deprecated.1
Agent configuration validation got a new CLI surface — agent experimental check-config and agent experimental onboard run a six-stage validation pipeline (file permissions, YAML syntax, API key format, site/region, live key validation, product summary) without requiring a running agent.2
LLM Observability span cost attribution arrived — dd-trace-py's LLMObs.annotate() now accepts a cost_tags argument to slice LLM spend metrics by team, project, or org dimension.3 The JS tracer added crash tracking for uncaught exceptions and MongoDB query value obfuscation as opt-in features. AppSec in dd-trace-js now always reports content-length and content-type response headers on enabled spans.4
Metrics are now queryable in plain English — Natural Language Queries in the metrics explorer let you describe what you want without writing query syntax.5
Toto 2.0 launched — a time series foundation model that scales reliably with compute, available in five open-weights sizes from 4M to 2.5B parameters.6
Database Monitoring added agentic investigation — the Database Investigator surfaces root causes and remediation steps for performance issues using an agent-based approach, with a new Supabase integration providing explain plans and one-click setup.78
Two named CVEs in the Docker SDK dependency (CVE-2026-34040, High; CVE-2026-33997, Medium) were patched in Agent 7.78.4 and 7.79.0.9 Agent 7.78.3 closed a PATH-hijacking CVE in the OTel Go SDK. Datadog for Government achieved FedRAMP High certification.10