microsoft-foundry

Deploy, evaluate, fine-tune, and manage Microsoft Foundry agents and models with azd.

Updated Jul 20, 2026
One-click install
npx skills add https://github.com/merceralex397-collab/alex-stack --skill microsoft-foundry-merceralex397-collab
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: microsoft-foundry
Source: https://github.com/merceralex397-collab/alex-stack/tree/main/dev/azure-profile/azure-skills-main/azure-skills-main/.github/plugins/azure-skills/skills/microsoft-foundry
Command: npx skills add https://github.com/merceralex397-collab/alex-stack --skill microsoft-foundry-merceralex397-collab

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires openai, azure-identity, azure-ai-projects, and includes scripts (resource) and references (resource) components.

What problem does it solve? Managing the full lifecycle of Microsoft Foundry agents and models requires juggling azd commands, MCP tools, RBAC roles, quotas, and evaluation pipelines. This Skill orchestrates those workflows end-to-end so developers can scaffold, deploy, evaluate, and fine-tune Foundry resources without memorizing each tool's syntax. ## Core Features & Use Cases - Agent Lifecycle Management: Scaffold, provision, deploy, invoke, schedule, and troubleshoot hosted and prompt agents through azd and Foundry MCP tools, with CI/CD pipeline setup. - Evaluation & Optimization: Run batch and continuous evaluations, curate datasets from production traces, optimize prompts, and run Agent Optimizer jobs on Python hosted agents. - Model Operations: Deploy models with preset, customized, or capacity-discovery routing, and fine-tune models using SFT, DPO, or RFT with grader calibration. - Use Case: A developer creates a new hosted agent with the quick-start workflow, deploys it to a Foundry project, runs batch evaluations against a trace-derived dataset, then fine-tunes the underlying model with DPO once quality regressions appear. ## Quick Start Ask the assistant to create and deploy a new hosted Foundry agent end-to-end using the azd quick-start workflow.

Frequently Asked Questions about microsoft-foundry

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I deploy a hosted agent to Microsoft Foundry?▼

Use the deploy sub-skill after reading the azd-guidance document. It covers deploying hosted agents with azd, smoke-testing the deployment, creating or updating prompt agents, and managing agent versions across multiple environments.

How do I fine-tune a model on Azure AI Foundry?▼

Use the finetuning sub-skill, which supports SFT, DPO, and RFT training types. It covers dataset preparation and validation, grader calibration for RFT, training job submission via scripts, checkpoint evaluation, and deployment of the fine-tuned model.

What is the difference between Basic and Standard Agent Setup in Foundry?▼

Basic setup is the default azd ai agent flow where all resources are Microsoft-managed with no capability host. Standard setup provisions a capability host with bring-your-own Cosmos DB, Azure Storage, and AI Search for data residency control, and takes 10-20 minutes to provision.

Why does my Foundry deployment fail with QuotaExceeded errors?▼

QuotaExceeded means your subscription's TPM or PTU allocation in that region is fully consumed. Check regional usage with az rest against the CognitiveServices usages API, delete unused deployments to free quota immediately, or request an increase through the Azure Portal.

Can I set up CI/CD for a Foundry hosted agent?▼

Yes, the cicd sub-skill covers setting up a CI/CD deployment pipeline for Foundry agents. You will typically need a service principal with at least the Foundry User role, created via az ad sp create-for-rbac, for pipeline authentication.

When should I not use the microsoft-foundry skill?▼

Do not use it for Azure Functions, App Service, or general Azure deployments, which belong to the azure-deploy skill, or for general Azure environment preparation, which belongs to azure-prepare. It is scoped specifically to Foundry resources, agents, models, and evaluations.