Microsoft runs 1,300 Azure SRE Agents that have mitigated tens of thousands of incidents across its own infrastructure. AWS shipped the same thing three weeks later. Both use MCP to reach each other's infrastructure without bespoke connectors. That is how a protocol becomes plumbing.
image from Gemini Imagen 4
Microsoft and AWS have both shipped production AI SRE agents, with Microsoft reporting 1,300+ agents mitigating 35,000+ incidents and saving 20,000+ engineering hours monthly, while AWS's DevOps Agent achieved a 77% MTTR reduction (2 hours to 28 minutes) at Western Governors University. AWS's Security Agent cuts penetration testing from weeks to hours, with Bamboo Health reporting findings no other tool surfaced. Both vendors have converged on MCP (Model Context Protocol) as the integration layer, enabling their agents to operate across competing cloud platforms and on-premises systems.
The autonomous operations race between AWS and Microsoft Azure is no longer theoretical. Both vendors have now shipped production AI agents with real deployment data, and the numbers tell a story that goes beyond the usual vendor benchmarking.
Azure SRE Agent reached general availability on March 10, 2026. Microsoft runs 1,300 or more agents internally against production systems and has mitigated more than 35,000 incidents with the platform — the GA post does not disclose a time frame for this figure. The same announcement states Microsoft saves more than 20,000 engineering hours each month. The agent operates with deep context — source code, logs, metrics, traces, infrastructure state, and tools — loaded at startup rather than fetched on demand. It builds a persistent knowledge base from every interaction and supports MCP connectors for ecosystem extensibility.
AWS DevOps Agent hit general availability the same window, also moving out of preview. Western Governors University documented a production investigation where the agent reduced total incident resolution time from an estimated two hours to 28 minutes — a 77% improvement in MTTR. AWS's own framing is that the DevOps Agent supports three to five times faster incident resolution. The AWS Security Agent runs autonomous penetration tests that Amy Herzog, AWS Vice President and CISO, described as cutting testing time from weeks to hours. Bamboo Health and HENNGE K.K. are named reference customers for the Security Agent, with Bamboo reporting findings that no other tool had surfaced.
The competitive positioning is deliberate. Both vendors are arguing that their agent is not a chatbot wrapper but an always-on infrastructure layer that replaces or augments human SREs. Azure calls its system an AI-powered operations teammate. AWS calls its set frontier agents — autonomous systems that work independently for hours or days, make decisions across multiple steps, and scale to concurrent tasks across an entire application portfolio. The language is different; the product category is the same.
What makes this more than a vendor comparison is the convergence on MCP as the integration layer. Azure SRE Agent explicitly supports MCP connectors. AWS DevOps Agent integrates with Azure and on-premises systems through the same protocol. Two cloud giants building agents that can reach across each other's boundaries is a signal that MCP is becoming the TCP of agent infrastructure — not because any single vendor designed it that way, but because the use case demands a standard nobody wants to own unilaterally.
The governance concerns are real and underreported. Both agents operate with significant autonomy against production systems. Neither vendor's announcement directly addresses what happens when the agent's reasoning about an incident is wrong, or how rollback works if an autonomous fix causes a different failure. The JetBrains warning about agentic coding repeating the cloud ROI crisis applies here: the speed of deployment is outpacing the governance frameworks that determine whether the autonomous agent is actually doing the right thing.
The competitive angle is the hook. The substantive story is the convergence: two major cloud vendors, independently, concluding that autonomous SRE agents are mature enough to bet production infrastructure on. That is a market signal worth noting even if you are not running either cloud.
Sources: Microsoft TechCommunity — Azure SRE Agent GA, AWS Machine Learning Blog — Frontier Agents GA, AWS DevOps Blog — WGU Case Study, Microsoft Learn — Azure SRE Agent Overview.
Story entered the newsroom
Research completed — 0 sources registered. Azure SRE Agent GA March 10 2026: 1300+ agents deployed internally, 35000+ incidents mitigated, 20000+ engineering hours saved/month. MCP connectors s
Draft (1179 words)
Reporter revised draft based on fact-check feedback (1174 words)
Reporter revised draft based on editorial feedback (1162 words)
Approved for publication
Headline selected: Microsoft Runs 1,300 AI Agents Against Production Systems
Published (556 words)
Get the best frontier systems analysis delivered weekly. No spam, no fluff.
Agentics · 2h 49m ago · 3 min read
Agentics · 5h 36m ago · 4 min read
