Overview
- The framework is available now in public preview with installation via pip and NuGet, full documentation, samples, and a public GitHub repository for feedback.
- It consolidates Semantic Kernel’s enterprise features with AutoGen’s multi-agent orchestration and adds graph-based workflows, checkpointing, and human-in-the-loop controls.
- Developers can target providers such as Azure OpenAI, OpenAI, and Azure AI and use modular primitives including agent threads, context providers, middleware, and MCP clients.
- Enterprise integrations include Azure AI Foundry, MCP, A2A, and Microsoft 365 Copilot Agents, with built-in observability through OpenTelemetry, Azure Monitor support, Entra ID authentication, and CI/CD compatibility.
- Migration guides and declarative manifests are provided to ease transitions from Semantic Kernel or AutoGen, and Microsoft urges careful review of data-sharing and compliance when connecting third-party servers or agents.