Complete pricing guide for Microsoft Foundry Agent Service. Compare all plans, analyze costs, and find the perfect tier for your needs.
Not sure if free is enough? See our Free vs Paid comparison →
Still deciding? Read our full verdict on whether Microsoft Foundry Agent Service is worth it →
mo
per use
annual
Pricing sourced from Microsoft Foundry Agent Service · Last verified March 2026
Azure AI Agent Service is now part of Microsoft Foundry (the platform previously branded as Azure AI Studio and Azure AI Foundry). Foundry is the umbrella platform for building AI applications and includes Foundry Models, Foundry Tools, Foundry IQ, and the Foundry Control Plane for observability. Agent Service is the specific component within Foundry dedicated to creating, deploying, and managing AI agents that can reason, use tools, and collaborate in multi-agent workflows.
You can access more than 11,000 models from the Foundry model catalog, including OpenAI GPT-4o and o-series, Meta Llama, Mistral, DeepSeek, and specialized industry and fine-tuned models. You can also bring your own fine-tuned or custom models and deploy them through the same managed infrastructure. The platform supports switching models at the configuration level without requiring code changes, and intelligent routing can automatically select the best model for each task.
Foundry Agent Service uses pay-as-you-go pricing based on the underlying resources an agent consumes: token-based model inference, storage, search queries (via Azure AI Search or Bing grounding), and compute for hosted agents. There is no separate per-agent fee. For higher-volume workloads, Microsoft offers Agent Commit Units (ACUs) as pre-purchased commitment plans that provide 5% to 15% discounts depending on the commitment level, applied across all resource types.
Yes. Multi-agent orchestration is a core capability, supported through Semantic Kernel and AutoGen integration as well as the native workflow agent type. You can define specialized agents (e.g., a researcher, a reviewer, and a writer) and coordinate them with patterns for delegation, handoff, sequential execution, group chat, and parallel processing. Human-in-the-loop approval steps can be inserted at any point in the orchestration graph.
Based on our analysis of the AI Agents category, Foundry trades some raw flexibility for enterprise managed-service benefits. OpenAI's Assistants API is simpler to start with but limited to OpenAI models and lacks Microsoft 365 integration, private VNet isolation, and Entra-based RBAC. Open-source frameworks like LangGraph, AutoGen, and CrewAI offer maximum customization and portability but require you to build and manage your own hosting, security, and observability infrastructure.
AI builders and operators use Microsoft Foundry Agent Service to streamline their workflow.
Try Microsoft Foundry Agent Service Now →The industry-standard framework for building production-ready LLM applications with comprehensive tool integration, agent orchestration, and enterprise observability through LangSmith.
Compare Pricing →Microsoft's open-source framework for building multi-agent AI systems with asynchronous, event-driven architecture.
Compare Pricing →Open-source Python framework for orchestrating role-playing, autonomous AI agents that collaborate as a 'crew' to complete complex tasks.
Compare Pricing →LangGraph is LangChain's open-source framework for building stateful, durable, multi-agent workflows in Python and JavaScript with graph-based control flow.
Compare Pricing →Developer platform for AI agent observability, debugging, and cost tracking with two-line SDK integration.
Compare Pricing →Dify is an open-source LLM app development platform that combines a visual workflow builder, RAG pipelines, agent tools, and an LLMOps backbone.
Compare Pricing →