Many of the world's largest enterprises run on .NET, and their teams are shipping AI agents just as fast as everyone else. With version 0.11 of our .NET AI SDK, you can now use AgentControl's core capabilities directly from your .NET applications, the same way Node.js and Python teams already do.

Run agents you control from LaunchDarkly

Your .NET applications can now pull agent configurations straight from AgentControl. Define your agent's instructions, model, and tools in LaunchDarkly, then change them at runtime without redeploying. Whether you are rolling out a new model, refining instructions, or targeting different behavior to different customers, your agents update instantly and safely.

Measure quality with judges

Judges let you evaluate the quality of your AI output using AI-powered evaluators that you define once and manage centrally in AgentControl. Your .NET applications can now use those judge definitions to evaluate responses in production and report the results back to LaunchDarkly, so quality shows up in your insights right alongside cost and latency, and your whole team works from the same definition of "good."

See what your agents are doing

New tracking capabilities capture latency, token usage, tool calls, and judge results with a single method call. That data flows into AI Insights, where you can compare cost, performance, and quality across models and variations.

The .NET AI SDK is in pre-release as we work toward full parity with our Node.js and Python AI SDKs. Try it from NuGet and check the docs to get started.