Hosted Agents in Microsoft Foundry is now Generally Available! This brings agent native compute for your agents with any framework, any language, and any model 🚀 Build with the GitHub Copilot App, connect to Microsoft IQ, run in Foundry, securely interact with Microsoft Teams, govern in Agent 365, and... optimize to compound your learnings continuously -- such a valuable combo. Check out the demo / blog below to see it in action!show more

this is really freaking good

Any framework, any language, any model is the real unlock.

Exposing agent‑native compute opens up low‑latency pipelines for any language. All the heavy lifting done here under the hood is just crazy. Especially at enterprise level.

The move toward agent-native compute reflects a broader shift in enterprise AI infrastructure. Flexibility across frameworks, models, and governance layers will be critical for organisations seeking reliable AI adoption at scale.

Thanks for sharing this, can i get more info on Evaluation feature

Any framework, any language, any model" on hosted agent compute sounds flexible. It also means any supply-chain compromise or context-poisoned workflow now runs with enterprise-grade resources and potential access to Teams/Microsoft services. Governance dashboards are nice, but they won't catch an agent that has legitimate reasons to take the harmful action because its inputs told it to. What's the enforcement model when the agent's reasoning is intact but the data is not?

Busy building them full throttle

Great walkthrough @jeffhollan. These are the type of core components that are going to help drive real adoption for an org.

Noticed in your demo you were using GPT 4.1 min not frontier modells how do you see different Model capabilities impact in this kind of work?

Portable across framework and model is the right pitch. What I'd want proven is whether the scheduler stays portable too. Every dispatch layer I've built on Aniso looked framework agnostic until the runtime's kernels leaked into job priority.