Craig Stanley
Home / Capabilities / Foundry / Foundry in plain terms

Foundry in plain terms

Microsoft Foundry is Azure's platform for building agents and apps on AI models. Here's what it is, how it differs from Copilot Studio, and when it fits.

· 3 min read · Craig Stanley
In short, explained

Foundry is a big workshop in Azure where builders pick an AI model, give it tools and instructions, and run it safely. It's for people who write software.

Microsoft Foundry is where developers build AI agents and apps on Azure. It gives access to a large catalogue of models, ways to build agents with or without code, and tools to test and monitor them. Copilot Studio is the low-code route inside Microsoft 365; Foundry is the developer route on Azure.

Foundry unifies agents (prompt, voice-based prompt, hosted), 10,000+ catalogue models and toolboxes (web/file search, code interpreter, MCP, SharePoint, Work IQ, Fabric IQ) under one Azure resource with Entra RBAC, networking and policy. Observability includes tracing and evaluations; built-in dashboards are preview. Agents can publish to Teams and Microsoft Copilot.

What it is

Microsoft's documentation says Foundry "unifies agents, models, and tools under a single management grouping" with built-in tracing, monitoring and evaluations. In plain terms, it's the Azure service where developers pick an AI model, give it instructions and tools, and run the result with enterprise controls around it.

Microsoft lists three things you can build with:

Building blockWhat Microsoft says it covers
AgentsPrompt agents and voice-based prompt agents built in the portal or SDK, and hosted agents that run your own code
ModelsMore than 10,000 models from Microsoft, OpenAI, Anthropic, Meta and others
Tools and knowledgeBuilt-in tools, memory and retrieval, grouped in a Foundry Toolbox

Two ways to build an agent

Microsoft describes a spectrum. At one end is a prompt agent: you choose a model, write instructions and attach tools, and Foundry runs it with no code or containers to maintain. At the other end is a hosted agent: you write the agent in a framework such as Microsoft Agent Framework or LangGraph, package it, and Foundry runs it with a managed endpoint, scaling and its own Microsoft Entra identity. Microsoft says you can start declarative and move to code as needs grow.

Tools include web search, file search and a code interpreter, plus custom tools through functions, OpenAPI specs and MCP servers. Microsoft's SDK documentation also lists SharePoint, Work IQ and Fabric IQ among the platform tools.

How it relates to Copilot

I think of three routes to an agent. Agent Builder is for anyone in Microsoft 365 who wants a simple agent. Copilot Studio is for makers who need actions, flows and more control with low code. Foundry is for developers who need a particular model, their own code, or an agent that runs inside an application.

The routes connect. Microsoft says Foundry agents can be published to Microsoft Teams and Microsoft Copilot, and Copilot Studio's billing page notes that Foundry models used through bring-your-own-model configurations are billed separately from Copilot Credits.

Why it matters for decisions

On this site I talk about decision models: a model that scores or classifies a case so a person or agent can act on it. If you want a specific model behind a decision, with your own thresholds and your own logging, Foundry is where that model is chosen, deployed and monitored. The rest of the experience can still live in Copilot. I sketch that pattern in Where a decision model sits in a Copilot agent.

A worked example

This scenario is illustrative. A lettings agency gets around 400 maintenance requests a month and wants each one sorted into "urgent", "routine" or "needs a call".

RouteFits?
Agent BuilderCould answer questions about the policy, but can't sort requests automatically
Copilot StudioCould sort requests with an agent flow and a prompt, metered in Copilot Credits
FoundryCould run a chosen model with fixed thresholds and full tracing, called from the agency's own system

If the agency's requests arrive in Outlook and staff work in Teams, I'd start with Copilot Studio. If they arrive in a property management system and need to be scored inside it, Foundry is the better fit. The choice follows where the work happens.

What I'm still checking

Foundry has changed quickly, including a new portal and a move from hub-based projects. Microsoft says some platform capabilities, such as built-in monitoring dashboards, are in preview and points to a general availability overview for current status. I'd check that page before relying on any specific feature in production.

Sources

Read next

A question to take awayWhich of these do you already pay for and not use?

About me

Craig Stanley

Microsoft AI consultant and technical architect, based in Whitley Bay. Over the last few years I've delivered Microsoft 365 Copilot, Copilot Studio agents, Microsoft Foundry (formerly Azure AI Foundry) work and governance for UK public sector and financial services organisations.

What interests me is the decision underneath the tool: what it costs, what it risks, and whether a small, transparent model can make it better. I write the methods up here and on Substack so anyone can use them.

I write this site to learn in public: explaining each idea simply is how I check I understand it. Why I write this site.

Find me