Foundry Agent Platform

Foundry Agent Platform Skill

Deploy, orchestrate, and manage AI agents on Microsoft Foundry — the unified Azure PaaS for enterprise AI.

Rapid Evolution Domain

Foundry is in active preview (February 2026). SDK versions change frequently.

Refresh triggers:

• azure-ai-projects SDK version bump (currently 2.0.0b3)
• Foundry portal feature releases
• Hosted Agents GA
• Memory API changes

Last validated: February 2026

Platform Overview

Microsoft Foundry (formerly Azure AI Foundry) unifies model hosting, agent orchestration, tool management, observability, and multi-channel publishing.

Foundry vs Other Platforms

Four SDK Types

This is the most common source of confusion. Foundry has four distinct SDK types, each with different endpoints and use cases:

SDK Packages

Breaking Change: Python 2.x is incompatible with 1.x. The 2.x uses .services.ai.azure.com endpoints.

Agent Service Patterns

Create Agent

from azure.ai.projects import AIProjectClient
from azure.identity import DefaultAzureCredential

client = AIProjectClient(
    endpoint="https://<resource>.services.ai.azure.com/api/projects/<project>",
    credential=DefaultAzureCredential()
)

agent = client.agents.create_agent(
    model="gpt-4.1-mini",
    name="my-agent",
    instructions="System prompt here."
)

Versioned Agents

from azure.ai.projects.models import PromptAgentDefinition

definition = PromptAgentDefinition(
    model="gpt-4.1-mini",
    instructions="System prompt",
    tools=[bing_tool, file_search_tool]
)

version = client.agents.create_agent_version(
    agent_id=agent.id,
    definition=definition
)

Conversations (Multi-Turn)

conversation = client.agents.create_conversation(agent_id=agent.id)

client.agents.create_message(
    conversation_id=conversation.id,
    role="user",
    content="Hello"
)

run = client.agents.create_run(
    conversation_id=conversation.id,
    agent_id=agent.id
)

Key Concepts

Tool Categories

File Search Setup

vector_store = client.agents.create_vector_store(name="knowledge")
client.agents.upload_file_and_poll(
    vector_store_id=vector_store.id,
    file_path="skills.pdf"
)
file_search_tool = FileSearchTool(vector_store_ids=[vector_store.id])

Memory & Foundry IQ

Memory is the cloud-native equivalent of Alex's synapse architecture — automatic, persistent, cross-surface.

Hosted Agents (Preview)

Containerized agents on managed infrastructure:

pip install azure-ai-agentserver-agentframework
agentserver run --interactive   # local test
agentserver run                 # container mode (port 8080)
azd deploy                      # deploy to Foundry

Supports any framework: LangGraph, MS Agent Framework, Semantic Kernel, custom.

Observability Stack

Agent → OpenTelemetry → Application Insights → Agent Dashboard

from azure.ai.agentserver import setup_observability
setup_observability(vs_code_extension_port=4319)  # local dev

Built-in Evaluators

Relevance, Groundedness, Coherence, Safety, F1, BLEU, ROUGE

Publishing Channels

One agent, many surfaces:

Realtime API (Voice)

Models: gpt-realtime (GA), gpt-realtime-mini (GA). Supports MCP tools during voice sessions, semantic VAD, image input. 30-min session limit, PCM16 mono 24kHz.

Authentication

from azure.identity import DefaultAzureCredential
credential = DefaultAzureCredential()  # Keyless (recommended)
client = AIProjectClient(endpoint=endpoint, credential=credential)

Anti-Patterns

Decision Checklist

When designing a Foundry-based agent:

• Which SDK type? (Foundry SDK for agents, OpenAI SDK for completions)
• Which model tier? (Premium for orchestrator, efficient for specialists)
• Agent Service or Hosted Agent? (Start with Agent Service; migrate later)
• What tools? (Bing, File Search, Code Interpreter, MCP, OpenAPI)
• Memory strategy? (Foundry Memory, File Search, or hybrid)
• Publishing targets? (API first, then Teams, then Web, then Voice)
• Evaluation plan? (Which evaluators, what dataset, what baseline)
• Auth model? (Entra ID keyless via DefaultAzureCredential)

Synapses

• [.github/skills/ai-agent-design/SKILL.md] (Critical, Implements, Bidirectional) - "Foundry is the runtime for agent design patterns"
• [.github/skills/multi-agent-orchestration/SKILL.md] (Critical, Implements, Bidirectional) - "Foundry Agent Service enables multi-agent orchestration at scale"
• [.github/skills/azure-architecture-patterns/SKILL.md] (High, Extends, Bidirectional) - "Foundry is an Azure PaaS that follows WAF principles"
• [.github/skills/mcp-development/SKILL.md] (High, Complements, Bidirectional) - "Foundry supports MCP servers as agent tools"
• [.github/skills/enterprise-integration/SKILL.md] (High, Enables, Forward) - "Foundry provides enterprise-grade agent deployment"
• [.github/instructions/meditation.instructions.md] (Medium, Created-During, Forward) - "Skill created during platform expansion meditation"

Microsoft Foundry Agent Platform — cloud-native agent deployment for the Alex ecosystem