Datadog

Recently shipped AI Guard for AWS Strands Agents and expanded RUM to .NET MAUI, watchOS, and visionOS.

AI Guard now protects AWS Strands Agents — a native Strands plugin evaluates prompts, model responses, and tool interactions at runtime, monitoring or blocking unsafe behavior without scattering security checks across application code. For Python users, the tracer added a per-provider kill switch (DD_AI_GUARD_OPENAI_ENABLED) and AI Guard support for LangChain 1.0+ agents.

RUM reached three new Apple platforms — .NET MAUI (via a single NuGet package), watchOS, and visionOS, all including crash reporting, error tracking, and network monitoring. The visionOS and watchOS support extends to GovCloud.

Bits AI gained Incident Response capabilities — Bits Investigation acts as an active AI responder to diagnose root causes, AI-generated chat summaries help responders catch up on active incidents, and bridge discussions are captured in a unified timeline. Bits Memories now reduce SAST false positives by evaluating findings against the application's full context, not just the flagged snippet.

Agent 7.81.0 switched metrics to the v3 intake by default — the more compact payload reduces outbound bandwidth. To keep using v2, set use_v3_api.series.enabled: "false". The Agent also removed the unsupported agent status py subcommand and renamed systemd units on Linux from datadog-agent-p to datadog-agent.

dd-trace-js v6.1.0 added support for @anthropic-ai/claude-agent-sdk and LLM Observability for Vercel AI SDK v7. Version 6.2.0 added GraphQL Mercurius support with a top-level graphql.request span. The Python tracer shipped Flask API endpoint discovery for every HTTP method (including auto-added HEAD and OPTIONS) and a new flask.resource.full tag for non-empty SCRIPT_NAME requests. LLM Observability now downgrades to the v0.4 trace API when APM and LLMObs are both enabled. The cluster agent also fixed a confused-deputy vulnerability in the AppSec admission mutator (7.80.2).