← EVENT LIBRARY / ai-agent

Microsoft Foundry turns production traces into an agent optimization loop

Microsoft expanded Foundry Agent Service with durable hosted execution, GA tool search, A2A and scheduled or event-driven Routines. Insights mines production traces for recurring failures and likely causes; a rubric evaluator, trace-to-dataset generation and Agent optimizer are scheduled to become GA later in September. Egress policies can run in audit mode and emit decisions to Application Insights. Voice agents and Insights remain preview features, while Microsoft's tool-token reductions are vendor-reported.

Imported from the existing research archive. Migration preserves the recorded judgment and is not a fresh verification.

What this means for you

For platform teams, agent operations are becoming a closed loop: observe production behavior, convert failures into evaluation data, test prompt, skill, tool or model changes, then deploy under network policy. Trial the GA pieces first and keep preview optimizers behind human promotion gates.

Evidence & related material

Background · Historical record, not rechecked

Microsoft expanded Foundry Agent Service with durable hosted execution, GA tool search, A2A and scheduled or event-driven Routines. Insights mines production traces for recurring failures and likely causes; a rubric evaluator, trace-to-dataset generation and Agent optimizer are scheduled to become GA later in September. Egress policies can run in audit mode and emit decisions to Application Insights. Voice agents and Insights remain preview features, while Microsoft's tool-token reductions are vendor-reported.

Technical details & scores
Ga tool search, A2A and scheduled/event-driven Routines
Preview voice agents, resilient hosted-agent features and Insights
Planned Ga rubric evaluator, trace-to-dataset generation and Agent optimizer later in September
Tool Search Claim >60% input-token reduction for 50 tools; >97% for 1,000 tools
Controls egress audit mode with decisions in Application Insights
Voice 80+ languages and 140+ locales
Technical impact 4
Engineering value 5
Adoption signal 4
Maturity 3
Verification cost 3
Risk 3