Once an AI agent hits production, the usual safeguards of code review and deployment gates no longer apply. LaunchDarkly's new AgentControl platform is designed to fill that gap, giving engineering teams the ability to change agent behaviour — including switching models or triggering fallbacks — in under 200 milliseconds, without a redeployment.
The platform launched on 19 May 2026 and marks a significant expansion of LaunchDarkly's feature-flag infrastructure into the AI agent layer. The governance problem it targets is straightforward: unlike traditional code, agent behaviour depends on prompts, model responses, and contextual inputs that vary in ways pre-deployment testing cannot fully capture. A configuration that looks stable in staging can produce inconsistent responses in production at scale.
AgentControl brings centralised oversight across teams and frameworks, live trace-level visibility into agent performance, and a benchmarking layer for progressively exposing changes before they reach live traffic. The 200-millisecond propagation window — inherited from LaunchDarkly's existing infrastructure — is fast enough to intervene within a single conversational turn.
The platform supports routing to a different model, triggering a fallback response, and versioning agent configurations across teams with no shared standard. Configuration changes propagate globally without touching underlying application code.
Cameron Etezadi, CTO of LaunchDarkly, said: "LaunchDarkly has always been about giving software teams control at runtime over what their software does in production. The hardest problems in AI, like model drift, unpredictable outputs, and the inability to intervene fast enough, turn out to be exactly the problems our platform was built to solve. We didn't have to reinvent the platform; we just had to extend it to meet the demands of an AI SDLC and agentic-driven workflows."
The launch comes with an endorsement from Cursor. Brian McCarthy, President, Global Revenue and Field Operations at Cursor, said: "Cursor is how the world's leading enterprises are building with AI. As more AI-powered products and agentic capabilities reach production, runtime control becomes essential infrastructure alongside the development workflows and controls teams already trust. LaunchDarkly built an additional layer for that environment, and AgentControl extends it to the agent lifecycle in a way that complements how Cursor's customers already build."
LaunchDarkly serves a quarter of the Fortune 500 and is headquartered in Oakland, California.
