Azure AI Foundry, now branded Microsoft Foundry, is Microsoft's enterprise platform for building, grounding and governing AI applications and agents. It gives access to more than 11,000 foundation, open, reasoning and multimodal models from providers such as OpenAI, Anthropic, Meta, Google, xAI and Hugging Face. Teams can fine-tune or distill models, let a model router pick the best option per task, host agents on serverless infrastructure and monitor them from a single control plane. Billing is consumption-based.
Software engineers, machine learning teams, platform architects and security or compliance leads at mid-size and large companies are the intended users. It suits organizations that already run on Azure and want agents to work with their existing business systems under consistent policy controls. Developers can start from templates, while governance staff use the control plane to manage every agent, tool and knowledge source.
Typical projects include building customer-facing assistants grounded in company data, automating business processes with action-oriented agents, and comparing several models before choosing one for production. Companies fine-tune a model on proprietary data, route requests between cheaper and stronger models to control cost, and run evaluations to track agent quality over time. Trust controls protect each agent, and observability tools show how it behaves in real time. Some workloads run at the edge, which helps where privacy or latency matters. Connections to Microsoft Fabric, Azure Machine Learning, Azure OpenAI and Azure Speech tie the work into the wider Azure ecosystem. Security, compliance and policy controls apply consistently across every agent, which matters for regulated industries that need audit trails and clear ownership of each deployed assistant.
