Microsoft Foundry & Foundry Agent Service Development
We build on Microsoft Foundry: choosing and deploying models, creating agents in Foundry Agent Service, running Microsoft Agent Framework code as hosted agents, and evaluating and tracing them before and after release.
Registered in Jodhpur since 2022
Dewiride Technologies Private Limited
Agent Framework tutorials
Step-by-step C# videos on our YouTube channel
Foundry setup guide
Creating a Foundry resource and deploying a model, on our developer blog
What is Microsoft Foundry?
Microsoft Foundry, formerly Azure AI Foundry, is Microsoft’s AI platform on Azure, which it describes as bringing agents, models and tools together under a single management grouping.
It covers a model catalogue (Foundry Models), Foundry Agent Service (formerly Azure AI Agent Service) for building and running agents, Foundry IQ for knowledge, and the evaluation, tracing and guardrails around them.
The new Foundry portal at ai.azure.com has been generally available since 16 March 2026 for core scenarios in Foundry projects, and Microsoft lists some capabilities as still in preview. Hub-based projects and standalone Azure OpenAI resources are available only in the older portal, now called Foundry (classic).
What we build on Foundry
Choosing and deploying models
We test candidate models from the Foundry Models catalogue on your own tasks and deploy the ones that fit. Non-Microsoft models include Anthropic’s Claude, available through an Azure Marketplace subscription: Anthropic sells and operates it and provides any SLA, and Foundry doesn’t apply its built-in content filtering to Claude at deployment, so we add our own checks.
Agents on Foundry Agent Service
We build prompt agents with the tools Microsoft lists as generally available in Foundry Agent Service, including MCP, OpenAPI, function calling, file search, code interpreter, web search and Azure AI Search. We flag preview tools, such as SharePoint grounding, before you rely on them.
Hosted agents for your own code
Hosted agents run your Python or C# agent code as a Foundry-managed service, with a managed endpoint, automatic scaling and a dedicated Microsoft Entra agent identity. Microsoft made them generally available in July 2026. The code can use Microsoft Agent Framework, LangGraph or another framework; we write ours with Agent Framework, whose Foundry hosting packages are still prerelease.
Knowledge with Foundry IQ
Foundry IQ is Foundry’s knowledge layer: permission-aware knowledge bases on Azure AI Search, with agentic retrieval, shareable across agents. Microsoft lists it as generally available at the API level, with portal access still in preview. We prepare and index the content behind it.
Evaluation and tracing
We run Foundry evaluations against your real questions before each release; Microsoft lists evaluations as generally available, with some evaluators in preview. Agent tracing follows OpenTelemetry conventions and stores traces in Azure Monitor Application Insights; it is generally available for prompt and hosted agents and in preview for workflow and external agents.
Agents in Teams and Microsoft 365 Copilot
Publishing Foundry agents to Microsoft 365 Copilot and Microsoft Teams is generally available, with documented limits such as no streaming responses or citations. When you need a fully customised agent there, we build it with the Microsoft 365 Agents SDK or Teams SDK.
Moving off classic agents and the Assistants API
The next-generation Foundry Agent Service, built on the Responses API, became generally available on 16 March 2026. Microsoft calls the agents that came before it classic agents.
| Service or feature | Retirement | Where Microsoft points you |
|---|---|---|
| Classic agents | Deprecated; scheduled to retire on 31 March 2027 | The new Foundry Agent Service |
| Azure OpenAI Assistants API | Retired; its documented sunset date was 26 August 2026 | Foundry Agent Service |
| Hosted agents deployed during the preview | The original hosting backend was supported only until 20 August 2026 | Redeploy on the new hosting backend |
| Foundry declarative workflows (preview) | Scheduled to retire on 1 December 2026 | Microsoft Agent Framework, run as a hosted agent |
| Prompt flow | Scheduled to retire on 20 April 2027 | Agent Framework |
Microsoft’s guide to migrating to the new Foundry Agent Service (opens in a new tab) covers classic agents and Assistants API code. Its migration tool helps automate Assistants API code changes but doesn’t move past threads or messages. Declarative workflows have their own workflow migration guide (opens in a new tab), and their retirement doesn’t affect code-based orchestration in hosted agents.
Take stock
We list every classic agent, Assistants API integration and tool, and what depends on them.
Map and rebuild
Following Microsoft’s mapping, threads become conversations, runs become responses and create_agent() becomes create_version(). Declarative workflows and Prompt flow logic become Agent Framework workflows in code.
Test side by side, then switch
We test each migrated agent against real conversations, alongside the original while it still runs, and switch over once it passes.
Security and governance
- Microsoft Entra agent identities: Foundry gives new agents their own identity by default, and hosted agents a dedicated one at deployment; we give each only the access its tools need. Entra Agent ID is generally available, and Microsoft states that extending Microsoft Entra security features to agents requires Microsoft Agent 365.
- Private networking for Foundry Agent Service, generally available since March 2026: your own virtual network, private endpoints and no public egress. Microsoft documents limits: traces and workflow agents don’t yet fully support network isolation; Grounding with Bing, web search and SharePoint grounding use public endpoints; hosted agents need virtual network injection set up when the Foundry account is created; and Microsoft’s deployment guide says a hosted agent’s endpoint stays publicly addressable, with each user’s sessions isolated by identity.
- Foundry guardrails (formerly content filters) use Azure AI Content Safety classifiers and cover models sold by Azure and agents built in Foundry Agent Service. They are generally available for model deployments and in preview for agents, including checks on tool calls and tool responses.
- Role-based access through Foundry’s built-in roles, such as Foundry User and Foundry Owner (the renamed Azure AI roles, with unchanged IDs and permissions). We give each person and service the narrowest role that works.
Foundry and Agent Framework together
Microsoft Agent Framework is Microsoft’s open-source SDK for building agents and multi-agent workflows in .NET and Python. Foundry Agent Service is Microsoft’s managed platform for building, deploying and scaling agents. Microsoft Learn calls Agent Framework “the recommended orchestration layer for Hosted agents in Foundry”.
We use them together in two ways: an Agent Framework app deployed as a hosted agent, or one in your own environment that connects to agents defined in Foundry. Neither requires the other: hosted agents run other frameworks, and Agent Framework runs outside Foundry too.
Agent Framework’s Microsoft Foundry connector is stable in Python but still prerelease in .NET. Our Microsoft Agent Framework development page covers its connectors and hosting options.
How we work and what it costs
Discovery call
A free call about your goals and Azure setup. If Copilot Studio, Microsoft’s low-code option, would do the job, we say so.
Pilot
A small pilot in your own Azure subscription, with agreed success criteria.
Production build
We set up identities, networking, guardrails, evaluations and tracing before go-live.
Handover and support
Code and documentation for your team, then support Monday to Friday, 9 AM to 6 PM IST.
Foundry usage is billed to your Azure subscription, separately from our fees. Hosted agents are billed for the CPU and memory used during active sessions, on top of model inference and tool usage.
- Consulting
- From ₹2,500 (US$40) an hour
- Monthly retainer
- From ₹2,00,000 (US$4,000) a month
- Pilots and fixed-scope projects
- Quoted per statement of work
GST is added to Indian invoices. Rate cards by business size and add-ons are on our pricing page.
Sources
- What is Microsoft Foundry? (opens in a new tab) — Microsoft Learn
- What is Microsoft Foundry Agent Service? (opens in a new tab) — Microsoft Learn
- Hosted agents in Foundry Agent Service (opens in a new tab) — Microsoft Learn
- Get started with Microsoft Foundry SDKs and endpoints (opens in a new tab) — Microsoft Learn
- Microsoft Foundry portal general availability overview (opens in a new tab) — Microsoft Learn
Last reviewed by Jagdish Kumawat, Founder & Director
Frequently Asked Questions
Is Microsoft Foundry the same as Azure AI Foundry?
Do we need Foundry to use Agent Framework?
What happens to our classic agents?
Can Foundry agents appear in Teams and Microsoft 365 Copilot?
Which models can we use?
How is Foundry billed?
What does your work cost?
Related services
- Microsoft Agent Framework DevelopmentWe design, build and run agents and multi-agent workflows on Microsoft Agent Framework, and move existing Semantic Kernel and AutoGen projects onto it.
- AI Agent DevelopmentAgents that call your systems, follow your rules and hand over to a person when they should, proved in a small pilot before they go to production.
- Custom AI SolutionsAI features built around your own data: retrieval-augmented generation (RAG), document processing and language-model features inside the software you already run.
Tell us what your agent should do
We'll tell you the simplest approach that will work — Microsoft Agent Framework, Copilot Studio or no agent at all — and what a pilot would involve.