Microsoft · Technologies
Foundry
Where your AI apps and agents get built.
Microsoft Foundry, formerly Azure AI Foundry, is the platform for building, deploying and operating AI apps and agents on Azure. One catalogue of frontier models, one set of tools and one governance surface. We design and build production AI on it.
FOUNDRY DEPLOYMENTS · EXAMPLE PROJECT
DEPLOYMENT
MODEL
PURPOSE
STATUS
Support-Agent
GPT-5.5
Customer email
LIVE
Doc-Summariser
Claude Opus
Contracts
LIVE
Router-v2
Model router
Cost control
TESTING
legacy-gpt4-app
Retired model
Unknown
MIGRATE
Insights-RAG
Foundry IQ
Knowledge
LIVE
Sample project
Live · testing · migrate
In plain terms
One platform for AI, instead of a different stack per project.
Foundry gives you one place to build, deploy and govern AI on Azure. Here is what changes.
Without Foundry
- Model choices scattered across tools, accounts and bills
- Every project builds its own scaffolding from scratch
- No single view of what is deployed and what it costs
- Model retirements discovered when something breaks
With Foundry
- One catalogue spanning OpenAI, Anthropic and more
- Shared tooling for agents, evaluation and safety
- One place to see and govern every deployment
- Model upgrades planned instead of survived
What CG TECH can do with Foundry
The work, broken into the parts that matter.
One catalogue, every frontier model
Foundry’s catalogue spans OpenAI, Anthropic’s Claude, Mistral, Meta and Microsoft’s own models, with a model router that picks the right model per request. You choose per workload instead of marrying one vendor.
the right model for each workload, not one default
Agents that run as a service
Foundry Agent Service, generally available since March 2026, runs agents with tools, memory and governance built in, so agents ship as managed services rather than science projects.
production agents, not demos
Answers grounded in your business
Foundry IQ handles retrieval across your enterprise knowledge and the live web, so apps and agents answer from your data rather than guessing.
answers grounded in your data
Run it like production software
Evaluation, monitoring, content safety and spend controls are part of the platform, and Agent 365 extends governance across your whole agent fleet.
AI you can put in front of auditors
How an engagement runs
From kickoff to production AI, step by step.
01
Discover
We map your use cases, data readiness and existing AI spend.
02
Design
We choose models, architecture and guardrails to fit the workload.
03
Build
We ship your first workload with evaluation and cost controls in place.
04
Handover
Your team gets the runbook, the dashboards and the training to own it.
Questions we hear a lot
Common questions about Foundry
What is Microsoft Foundry?
Foundry, previously called Azure AI Foundry, is Microsoft’s platform for building, deploying and running AI applications and agents on Azure. It gives you one catalogue of models, one set of build tools and one place to govern what you have built.
Is Foundry the same as Azure OpenAI?
Azure OpenAI is one part of it. Foundry is the broader platform, renamed from Azure AI Foundry, covering the full model catalogue, agents, knowledge and governance.
Which models can we use?
The catalogue includes OpenAI's models, Anthropic's Claude, Mistral, Meta and Microsoft's own, with more arriving steadily. A model router can pick the best model per request.
What does Foundry cost?
It is consumption based Azure pricing, so cost depends on the models and volume you use. We model expected costs during discovery so there are no surprises.
We already built agents in Copilot Studio. Do we need Foundry?
They complement each other. Copilot Studio suits business built agents inside Microsoft 365. Foundry suits custom, pro code apps and agents. Many estates run both.
Can we use models other than Microsoft’s?
Yes. Foundry offers a catalogue of frontier models rather than one vendor’s, and Anthropic’s Claude models are generally available in it. Being able to pick a model per job, and change your mind later, is a large part of why businesses build there.
When should we build in Foundry rather than buying Copilot?
Buy first, build second. If Microsoft 365 Copilot or a Copilot Studio agent does the job, that is cheaper and faster. Foundry is for the work those cannot do: your own application, unusual data, or a model choice the packaged products do not offer.
What skills do we need in house?
Less than people expect to start, more than people expect to keep running. A first build can be delivered by us. Running it afterwards needs someone who owns it, and we would rather set that up with you than have you discover the gap later.
Ready when you are
Building AI on Azure? Let us talk.
A discovery session maps your use cases, your risks and your quick wins. You keep the plan either way.
What to expect
- A consultant replies within 4 business hours
- Session booked to understand your requirements
- We will provide you with a fixed price quote