Azure Functions Ships Serverless Agents Runtime at Build 2026

2026-06-19 · Source: InfoQ · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure, Software Development & Engineering · Depth: Intermediate, short

Summary

Microsoft introduced the Azure Functions serverless agents runtime in public preview at Build 2026 on June 19, 2026, transforming the event-driven compute service into a platform for hosting AI agents. This new runtime defines agents using ".agent.md" files, a markdown-first programming model that consolidates an agent's instructions, tools, connections, and behavior into a single document. Agents can be triggered by various Azure Functions triggers, including HTTP, Timer, and new connection-backed triggers for Microsoft 365 services. They access MCP tool servers, sandboxed code, browser execution via Azure Container Apps, and over 1,400 managed connectors. The operational model maintains existing Flex Consumption benefits like scale-to-zero and per-second billing, with no additional cold start or "agents tax." Other Build 2026 announcements included the General Availability of the MCP extension and Flex Consumption rolling updates, significant Durable Task Scheduler enhancements, and Go language support.

Key takeaway

For AI Engineers building enterprise-grade AI agents, Azure Functions' new serverless agents runtime offers a compelling platform. You can now define complex agents using a markdown-first ".agent.md" model, leveraging existing serverless benefits like scale-to-zero and per-second billing without an "agents tax." This simplifies agent development and deployment, allowing you to focus on agent logic and tool integration rather than infrastructure. Explore the public preview to streamline your agent orchestration.

Key insights

Azure Functions now offers a serverless runtime for AI agents, simplifying development with a markdown-first programming model.

Principles

• Agent definitions can be markdown-centric.
• Serverless platforms can host complex AI agents.
• Agent execution cost is tied to LLM invocation.

Method

Agents are defined in ".agent.md" files with YAML frontmatter for triggers/metadata and markdown for instructions. Companion files declare tools and MCP servers, with the runtime handling orchestration.

In practice

• Define agents using the ".agent.md" format.
• Integrate agents with 1,400+ enterprise connectors.
• Automate security posture audits with timer-triggered agents.

Topics

• Azure Functions
• Serverless AI Agents
• .agent.md
• MCP Extension
• Flex Consumption
• Durable Task Scheduler

Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Engineer, Machine Learning Engineer, Software Engineer

Related on AIssential

• Build highly scalable serverless LangGraph multi-agent systems in AWS with Amazon Bedrock AgentCore — LangGraph, AWS serverless, and Amazon Bedrock AgentCore enable scalable, observable, and stateful multi-agent generative AI systems.
• How Headless Agents Will Change Work — The shift to headless software and agent-first platforms is fundamentally altering enterprise software consumption and business models.
• Claude Managed Agents: get to production 10x faster — Claude Managed Agents simplify AI agent deployment by abstracting infrastructure, accelerating development from months to days.
• Azure Logic Apps Adds Sandboxed Code Interpreters to Agent Workflows — Azure Logic Apps now integrates sandboxed code interpreters for secure, AI-driven data analysis and transformation within workflows.
• Boost agent reliability with W&B Training Serverless RL — W&B Training Serverless RL simplifies LLM fine-tuning via reinforcement learning with managed GPU capacity and observability.
• Daytona raises $24M Series A to build agent-native compute infrastructure — Agent-native compute infrastructure is crucial for scaling stateful, long-running AI agent workloads beyond traditional cloud models.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by InfoQ.