Microsoft Foundry Prompt Optimizer
learn.microsoft.com
Spec sheet
- Purpose
- Prompt Optimizer automatically improves an agent’s system instructions using prompt-engineering best practices.Read 4 Oct 2026 · learn.microsoft.com
- Headquarters
- Redmond, Washington, United StatesRead 4 Oct 2026 · learn.microsoft.com
- Founded
- 1975Read 4 Oct 2026 · learn.microsoft.com
- Maker
- Microsoft’s corporate headquarters are in Redmond, Washington.Read 4 Oct 2026 · news.microsoft.com
- Access
- The optimizer is available in the Foundry portal’s agent configuration and in the model playground wherever a system-instructions field is present.Read 4 Oct 2026 · learn.microsoft.com
- How it works
- An LLM uses the supplied instructions and optional suggestions to restructure and clarify the prompt.Read 4 Oct 2026 · learn.microsoft.com
- Iteration
- Users can add suggestions and reoptimize, using the latest optimized text as the next baseline.Read 4 Oct 2026 · learn.microsoft.com
- Limits
- It optimizes text-based agent instructions and does not optimize non-text configuration such as tool definitions or knowledge sources.Read 4 Oct 2026 · learn.microsoft.com
- Persistence
- Optimization results are not stored permanently, and the optimizer does not automatically save previous instruction versions.Read 4 Oct 2026 · learn.microsoft.com
- Prerequisites
- Use requires a Foundry project with a prompt or workflow agent, a model deployment in a supported region, and access to the agent configuration panel.Read 4 Oct 2026 · learn.microsoft.com
- Preview status
- The feature is in public preview, without a service-level agreement, and Microsoft does not recommend it for production workloads.Read 4 Oct 2026 · learn.microsoft.com
- Reasoning
- It provides an explanation for each modified paragraph that users can review alongside the revised text.Read 4 Oct 2026 · learn.microsoft.com
- Regions
- Supported project regions listed include Central US, East US 2, France Central, Germany West Central, Italy North, Japan West, North Central US, Poland Central, Spain Central, Sweden Central, Switzerland West, UAE North, West US, West US 2, and West US 3.Read 4 Oct 2026 · learn.microsoft.com
- Related optimization
- The separate Agent Optimizer can generate candidates for prompt-agent instructions, function-calling tool descriptions, or model selection.Read 4 Oct 2026 · learn.microsoft.com
- Security and privacy
- Microsoft states that prompts and completions for Models sold by Azure are not shared with other customers or used by model providers to improve their models or services.Read 4 Oct 2026 · learn.microsoft.com
Where Specifiction places it
- AI Prompt Generators#24 of 286.8