Microsoft Foundry development
Microsoft Foundry Development Services
Build, evaluate and run AI agents and generative AI applications on Microsoft Azure.
Microsoft Foundry (formerly Azure AI Foundry) is Microsoft's platform for building AI applications and agents on Azure. We help you choose models, ground them in your data, add guardrails and evaluation, and connect the result to the systems your teams already use.
What we cover
- Model selection and prompt design
- Retrieval over your own data (RAG)
- Agents with tools and actions
- Content safety and guardrails
- Evaluation and quality testing
- Tracing, monitoring and cost
Requirement-led advice · Documented handover
- Advice based on your requirements
- Works with your existing stack
- Documented, maintainable delivery
What Is Microsoft Foundry?
Microsoft Foundry is an Azure service that brings together a catalogue of AI models, an agent service, evaluation tools, safety controls and monitoring in one place. Teams use it to build chat assistants, document-processing workflows and agents that call business systems, while keeping data, identity and networking inside their Azure environment.
The model catalogue includes OpenAI models offered through Azure as well as models from other providers, so the choice of model can follow the task, cost and data-residency needs rather than a single vendor. Projects in Foundry group the models, data connections, agents and evaluations for one application, which makes it easier to separate development, testing and production.
When Microsoft Foundry is a good fit
- Your data, identity and security controls already live in Microsoft Azure and Microsoft Entra ID.
- You need more than one model, or want to compare models before committing.
- You want evaluation, content safety and tracing built into the release process.
- The AI feature has to call internal APIs, databases or Microsoft 365 data.
Where Teams Use Microsoft Foundry
Typical applications. Each starts with a clear task and a way to measure whether the answers are good enough.
Internal knowledge assistants
Answer staff questions from policies, manuals and project documents, with sources shown.
Document processing
Extract, classify and summarise contracts, invoices and reports.
Task agents
Agents that look up records, draft responses or create tickets through approved tools.
Customer support drafting
Suggest replies grounded in product documentation for agents to review.
Enterprise search
Natural-language search across documents and structured data.
AI features in your product
Add summarisation, extraction or chat to an existing web application.
Data question answering
Let users ask questions about reports with answers tied to the data.
Policy and compliance checks
Flag missing clauses or policy issues for human review.
Microsoft Foundry Services We Deliver
Each piece of work has a clear output, so you know what you will receive.
Use-case assessment
Shortlist AI use cases, check data readiness and estimate effort and running cost.
Deliverables: prioritised use cases, data review, cost estimate
Proof of concept
A working prototype on your data with measured answer quality.
Deliverables: prototype, evaluation results, go/no-go recommendation
Retrieval (RAG) pipelines
Index documents with Azure AI Search and keep answers grounded in your sources.
Deliverables: indexing pipeline, retrieval design, citation handling
Agent development
Agents with defined tools, permissions and human approval steps where needed.
Deliverables: agent design, tool integrations, test scenarios
Evaluation and safety
Test sets, automated evaluations and content safety settings before release.
Deliverables: evaluation suite, safety configuration, release checklist
Production operations
Deployment, tracing, monitoring, quota and cost control after go-live.
Deliverables: deployment pipeline, dashboards, runbook
Microsoft Foundry Building Blocks
The parts of the platform we most often work with.
Models
- Azure OpenAI models
- Other catalogue models
- Fine-tuning where justified
- Model comparison and benchmarks
Agents and apps
- Foundry Agent Service
- Tool and function calling
- Prompt and workflow design
- SDKs for Python and .NET
Data and grounding
- Azure AI Search
- Azure Blob Storage
- Microsoft Fabric and SQL data
- Connectors to business systems
Trust and operations
- Azure AI Content Safety
- Evaluations
- Tracing and Application Insights
- Microsoft Entra ID and private networking
Listing a feature here does not mean every project uses it. We recommend what fits your requirements, skills, licensing and running costs.
How Foundry Fits Your Stack
Foundry applications usually sit next to systems you already run. We build the application layer in .NET and ASP.NET Core or Python, host it on Microsoft Azure, and ground answers in data from Microsoft Fabric, SQL Server and Azure SQL or SharePoint.
For business users who want to build agents without code, Microsoft Copilot Studio can sit in front of the same data. For complex multi-step agents, frameworks such as LangChain and LangGraph can run against Foundry models, and Model Context Protocol servers can expose your tools to them.
Implementation Considerations
- Model availability, quotas and pricing vary by Azure region, so we check them before choosing a design.
- Answer quality depends on the source documents; we review and clean the content that grounds the answers.
- Agents should have the least access they need, with approvals for actions that change data.
- Evaluate before release and keep evaluating, because model updates can change behaviour.
- Plan for token cost and latency from the start, including caching and smaller models where they are good enough.
- Keep a person in the loop for decisions with legal, financial or safety impact.
How We Work With Microsoft Foundry
The same four steps we use across our technology work, adapted to your scope.
Discover
Goals, users, constraints and the systems Microsoft Foundry must work with.
Select
Confirm Microsoft Foundry is the right fit and agree the architecture and scope.
Build and integrate
Deliver in short iterations with testing, reviews and documentation.
Operate and improve
Monitor, support and improve the solution as your needs change.
Microsoft Foundry Questions, Answered
What is Microsoft Foundry?
Microsoft Foundry is Microsoft's Azure platform for building AI applications and agents. It combines a model catalogue, an agent service, evaluation, content safety and monitoring.
Is Microsoft Foundry the same as Azure AI Foundry?
Yes. Microsoft renamed Azure AI Foundry to Microsoft Foundry. Existing Azure AI Foundry projects continue in the renamed service.
Which AI models can we use in Microsoft Foundry?
OpenAI models offered through Azure and models from other providers in the Foundry catalogue. Availability depends on region and your subscription, so we confirm it for your case.
Does our data leave our Azure environment?
Foundry runs inside your Azure subscription and supports Microsoft Entra ID and private networking. We review the data flows and the terms for each model you plan to use.
How do you measure answer quality?
We build a test set from real questions, run automated evaluations and review samples with your subject experts before release.
Should we use Foundry or Copilot Studio?
Copilot Studio suits low-code agents inside Microsoft 365. Foundry suits custom applications, advanced retrieval and agents that need full developer control. Some solutions use both.
Planning a Microsoft Foundry Project?
Tell us the task, the data involved and the systems it must work with. We will suggest a practical first step.
Product names and logos belong to their respective owners. Mentioning a technology does not imply a partnership, certification or endorsement.