Harness adds CI/CD pipelines, governance and observability to AI agents, letting teams build, test, deploy and secure them like any other software. New tools, AgentTrace, AI Evals, AI Asset Catalog and an AI firewall, address regression testing, decision‑path debugging and prompt‑injection risks. This pushes AI agent management into the DevOps mainstream.
GitLab Duo’s new “Work item created” trigger fires the moment an issue opens, automatically routing it based on team capacity. The flow runs in seconds, scaling from a single ticket to hundreds, and eliminates the manual triage bottleneck that slows delivery.
Datadog now supports native OTLP ingestion via the standard OTLP HTTP exporter and a direct endpoint, letting you send traces, metrics, logs without any Datadog-specific agents. This lets teams keep a vendor‑neutral telemetry pipeline while still using Datadog’s APM and infrastructure tools for investigation.
Deploying Postgres, Redis or OpenSearch on Kubernetes is a few commands away, but the hidden cost lies in upgrades, backups, patching and failure recovery. A centralized platform that automates the full database lifecycle lets developers request services through K8s while hiding operational complexity, turning a hard problem into a sustainable service.
The Cloud Cost skill integrates cost and observability data into Datadog’s Bits Chat, letting teams ask plain‑language questions about cloud, AI, and SaaS spend. It can identify anomalies, trace spend to teams or services, and correlate cost changes with metrics, speeding up FinOps investigations without swapping dashboards.
Amazon Bedrock AgentCore, Microsoft Foundry, and Gemini Enterprise Agent Platform now share a common stack, runtime, memory, tool gateway, identity, observability, and governance. This convergence signals a de‑facto standard for enterprise AI agents, making cross‑cloud portability and vendor lock‑in harder and pushing the market toward a unified contract layer.
An AI agent mistakenly billed $4,000 to the wrong account because it ran with broad API credentials and no infrastructure checks. Red Hat AI’s BYOA platform adds identity, sandboxing, and scoped credentials across any agent framework, stopping such costly errors. The approach protects existing framework investments while delivering production‑grade security.
HashiCorp’s new tfpolicy lets platform teams write governance rules directly in HCL, eliminating the need for external tools like Sentinel or OPA. Built into Terraform and available in public beta on HCP Terraform, it can evaluate resource relationships, provider usage, and live state, tightening compliance without a separate policy stack.
Red Hat’s new MaaS governance layer lets admins set token‑based consumption limits (MaaSSubscription) and fine‑grained model access rules (MaaSAuthPolicy). This dual‑gate approach ensures teams only use authorized models while staying within quota, simplifying enterprise AI cost control and security.
You’re caught up.
That was all 9 stories in the 21 July issue.