Dapr 1.19 Improves Workflow Performance and Adds AI Agent Capabilities
The latest Dapr release delivers significant throughput and latency optimizations for workflows while introducing new governance policies and alpha APIs for AI agent workloads.
Dapr 1.19 dropped on October 22, 2026, bringing workflow throughput increases of up to 40% and a 71% drop in p99 latency for actor reminder registration. These figures come straight from nightly performance tests comparing this version against 1.18.0. P99 latency for job scheduling also fell by 33%. As a graduated CNCF open source project pulling over 700,000 weekly NuGet downloads, the runtime touches several core areas.
If you write in .NET, the SDK now sends only new events from each turn to workflow workers instead of the full execution history, reducing overhead. You also get new operational tools to list, inspect, and rerun workflows.
Version 1.19 introduces WorkflowAccessPolicy for governance. This lets an application operate on workflows owned by other apps while controlling permitted operations. You can configure "requires" rules that restrict an operation unless the caller's history contains specific listed events.
For AI agent workloads, the release adds alpha APIs for file storage, document search, and vector queries, alongside a Dapr MCP server. Operations get a simpler control plane because actor placement can now run directly from the Scheduler, letting you drop one service.