Spillwave Solutions Spillwave Guides

Azure Microsoft Foundry: Deploy your Harness to the Hyperscaler Control Plane

Your agent loop works. It runs on your laptop, or in a container your team operates, and it does the job. The open question is a different one: when your organization standardizes on Azure, what does the platform own, and what do you still get paged for at three in the morning?

Azure Microsoft Foundry: Deploy your Harness to the Hyperscaler Control Plane answers that by taking one real agent, a research desk serving many analysts, from a laptop loop to a governed, observable, multi-tenant fleet. The through-line is a single claim: the loop is yours, the substrate is theirs. Foundry supplies hosted compute, a protocol contract, a governed tool registry, managed conversation and session state, durable state, memory, agent identity, guardrails, tracing, evaluation, and a fleet control plane. You keep the loop.

The series runs two frameworks on one substrate, because Foundry's own documentation runs two. Microsoft Agent Framework is the native path. LangChain DeepAgents and LangGraph is the bring-your-own path. Every platform part shows both wirings, or says plainly why a feature is native-only.

Every factual claim is checked against Microsoft's own documentation mirror, which matters here because that documentation disagrees with itself about catalog sizes, general-availability dates, and which features survived the rename from Azure AI Studio to Azure AI Foundry to Microsoft Foundry. Where the pages contradict each other, the guide says so and names both. Seventeen parts and four appendices.

Start at Part 1, or jump to any part below.

The parts

  1. Microsoft Foundry Is Not a Product You Log Into. It Is an ARM Resource.
  2. Your agent is four lines away from a governed cloud endpoint
  3. Every platform says it is Framework-Agnostic. here is the receipt.
  4. Your Foundry agent is deployed and has no idea who is calling it
  5. There is no FoundryCheckpointSaver
  6. Your agent's tools should not ship on your release cadence
  7. Microsoft Foundry Ships a Thousand Tools. The Only Two Questions That Matter Are Not on the List.
  8. Foundry memory does the expensive part, then hands you the harder one
  9. One agent, three hundred users, one endpoint, and a security reviewer who wants proof
  10. Your agent cannot start itself, survive a hangup, or survive a crash, and those are three different bugs
  11. Your agent pays frontier prices to summarize a press release
  12. The Paragraph in the PDF Was Addressed to Your Agent
  13. Foundry Is Not Tracing Your Agent, and Nothing Will Tell You
  14. Your agent is fully traced and still ungradeable
  15. Microsoft Foundry Multi-Agent: The Protocol Is a Weekend. The Permissions Are the Project.
  16. Your agent fleet dashboard is lying to you, and the permission model is why
  17. The Azure AI agent decisions you cannot undo
  18. The publish button is four API calls in a trench coat
  19. Hosted Agents Do Not Exist in Azure Government. Read That Table Before the ATO Review.
  20. Microsoft Foundry Lock-In Is Real. It Is Not Where You Think It Is.
  21. Two Control Planes, Nine Axes, and the Third Column This Article Refuses to Print