Azure SRE Agent: general availability of Microsoft's operations agent

2026-03-10 · Publisher : Microsoft · Domain : Cloud & DevOps

On 10 March 2026, Microsoft declares the Azure SRE Agent generally available: more than 1,300 agents deployed internally, 35,000 incidents mitigated and over 20,000 engineering hours saved, according to Microsoft.

What changes

  • Azure SRE Agent is generally available on 10 March 2026.
  • Microsoft figures: 1,300+ agents, 35,000+ incidents mitigated, 20,000+ hours saved.
  • Autonomy is adjustable, from recommendation to guarded automated response.
  • Extensions through skills, subagents, Python tools and MCP connectors.

### What changes On 10 March 2026, Microsoft announces general availability of Azure SRE Agent, an AI-powered "operations teammate" that helps diagnose and resolve production incidents. Microsoft says it uses it on its own services: more than 1,300 agents deployed, over 35,000 incidents mitigated and over 20,000 engineering hours saved. The tool was previously in preview. ### Key facts - The agent works with source code, logs, metrics, traces, infrastructure state and the team's tools, loaded in advance. - A code interpreter runs Python to produce deliverables: PDF reports, charts, Excel workbooks, dashboards. - Memory and learning: the agent builds its knowledge base from every incident, and the team can also teach it directly. - The level of autonomy is adjustable, from simple recommendations to automated responses within defined limits. - Extensions: skills, subagents, Python tools, hooks, MCP connectors. Billing moves to a per-token measure on 15 April 2026. ### Why it matters After AWS, which presented its agent in preview in December 2025, Microsoft confirms that the operations agent is becoming a product category. The figures cited are the vendor's own, and the announcement does not detail how they are computed, but they signal an ambition to cut resolution time and on-call load. For SRE teams and consulting firms running production, the value will depend on the quality of the context supplied and the level of autonomy accepted. ### Takeaway Agent-assisted on-call becomes a standard cloud product; trust is built on guardrails and traceability.

Links

Tags : Agents, Observabilité, Azure, MCP, IA