Agents Talking to Agents (A2A)

As enterprises scale their assistants, they often face a tough reality: many critical capabilities live in third-party agents or legacy systems that can’t be refactored into one monolithic assistant. This is where Agent-to-Agent (A2A) comes in.

A2A provides a standard way to plug external agents into assistants as “black boxes” while still keeping the conversation smooth, contextual, and coherent. In this session, we’ll break down how Rasa uses A2A to let your assistant borrow skills from external specialized agents, when this approach makes sense (and when it doesn’t), and what design patterns actually work in production.

You’ll see real examples from our “Ace” demo assistant that blends native flows, MCP-powered tools, and A2A-integrated services into a single seamless conversation.

What you’ll walk away with

• A clear understanding of what A2A is (and isn’t)
• When to use A2A vs. building skills directly into your assistant
• Common orchestration patterns for integrating black-box agents
• How Rasa handles A2A tasks, context passing, and conversation control
• Lessons from real-world use cases: where A2A reduces complexity, and where it introduces it

Stay tuned for the third installment of the Rasa Orchestration Series.