Dapr’s microservices runtime now supports AI agents

In 2019, Microsoft introduced Dapr, an open-source runtime designed to simplify the development of distributed microservice-based applications. While AI agents were not a prominent topic at the time, Dapr included key components that laid the foundation for supporting AI agents. This was made possible through Dapr’s virtual actors feature, allowing for independent message processing within the system.

Today, the Dapr team has launched Dapr Agents, offering developers a comprehensive set of tools to build AI agents efficiently.

Dapr Agents evolved from Floki, an open-source project that extended Dapr for AI agent applications. By joining forces with Microsoft AI researcher Roberto Rodriguez and other project maintainers, Dapr ensures the seamless integration of the new agent framework.

According to Dapr co-creator Mark Fussell, agentic systems represent a new perspective on distributed systems, offering enhanced capabilities for language models and other functionalities.

Dapr Agents leverage Dapr’s efficient actors for seamless coordination and state management. These actors can quickly process messages and preserve their state, contributing to resource efficiency.

Currently, Dapr Agents support various model providers such as AWS Bedrock, OpenAI, Anthropic, Mistral, and Hugging Face, with local LLM support coming soon.

In addition to model interactions, Dapr Agents allow developers to define a list of tools for agents to utilize in completing tasks efficiently.

Python is currently supported by Dapr Agents, with .NET support on the horizon and plans for Java, JavaScript, and Go support in the near future.