Microsoft Agent Framework

In this catalog it is the Workflow Orchestration option when you need to run durable multi-step jobs with retries and state. Keep Microsoft Agent Framework on the observe/propose side of the loop: useful signals and drafts, with human or policy authority still required before anything irreversible runs.

Workflows are coded or declared; the engine persists state, retries activities, and wakes on signals/timers until completion.

1. Declare the workflow/graph and activity boundaries.
2. Configure durable state / event history storage.
3. Run a failing activity and confirm retry/resume behavior.
4. Expose signals/timers the agent host will use in production.

Related: semantic-kernel · autogen · topics/01-architecture · topics/12-central-nervous-system

• Use when: Microsoft Agent Framework is a production-oriented Python and matches the bottleneck and you specifically need to run durable multi-step jobs with retries and state.
• Skip when: another pack already covers the need with less ops surface, or Microsoft Agent Framework would become a second place people retype truth.
• Prefer instead: compare Semantic Kernel, Autogen before locking the cell.

• Ops: confirm license, hosting, and release cadence for Microsoft Agent Framework (microsoft/agent-framework) before relying on it in a live corridor.
• Integration: bad credentials, scopes, or schema drift produce empty retrievals or runaway tool/model spend without an obvious UI alarm.
• Scope: Workflow Orchestration components support agents; they are not the company ledger or the approval authority.