Cannot use MAF 1.2.0 with azure-ai-agentserver-agentframework — pinned to <=1.0.0rc3

[Arturo-Quiroga-MSFT]

Heads-up for anyone hosting MAF agents in Azure AI Foundry via the agent server adapter.

The latest azure-ai-agentserver-agentframework (1.0.0b17, 2026-03-16) hard-pins:

Requires-Dist: agent-framework-azure-ai <=1.0.0rc3,>=1.0.0rc2
Requires-Dist: agent-framework-core <=1.0.0rc3,>=1.0.0rc2

So from_agent_framework(my_agent).run() is effectively stuck on the pre-release rc3
line. Stable MAF (1.0.0 → today's 1.2.0) cannot be installed alongside it — pip either
downgrades MAF back to rc3 or fails resolution outright.

Practical impact: none of the GA fixes / features (Foundry hosted agent V2, Toolboxes,
checkpoint security hardening, OTel improvements, etc.) are reachable from a hosted-agent
deployment today.

I filed an issue against the SDK repo where the adapter lives:
👉 Azure/azure-sdk-for-python#46521

Posting here for visibility in case other folks are hitting the same wall, and to ask
the MAF maintainers: is there a coordinated release planned that bumps the adapter's
ceiling to track MAF stable (e.g. >=1.0.0,<2)?

Workaround until then: pin everything to the rc3 / b17 combo and don't let pip drift.

[moonbox3]

@Arturo-Quiroga-MSFT please follow the latest instructions here - the hosting scenario for Foundry agents was recently re-written to bring in hosted agents updates. https://github.com/microsoft/agent-framework/tree/main/python/samples/04-hosting/foundry-hosted-agents