Agents across platforms (A2A)
A2A (Agent2Agent) is an open standard that lets agents from different platforms work together: one agent hands a task to another, the other does it and sends back the result. Each agent publishes a card that says who it is, what it can do and how to talk to it.
In NappAI, A2A works both ways:
| Direction | What you get | Page |
|---|---|---|
| Publish your agent | A NappAI flow becomes an A2A agent that other agents, in NappAI or on other platforms, hand tasks to. | Publish an A2A agent |
| Connect to another agent | Your agents hand work to an A2A agent, whether it is in NappAI or elsewhere. | Connect to another agent |
This is how both are done:
What you see in the video
- A travel planner flow, which prepares itineraries and budgets, is published by adding the A2A Server component, with its name, description and skills.
- After deploying a version, the A2A Server tab of API Access shows the agent’s public card, its address and examples to call it.
- A travel agency’s assistant gets the A2A Agent component, with the planner’s address and a credential holding the key.
- In the playground, a customer asks for a trip to Lisbon: the assistant hands the plan to the planner over A2A and presents it to the customer.
A2A or MCP
Both are open standards, but they serve different purposes:
- With MCP an agent uses tools: specific functions, such as checking a domain or creating a task.
- With A2A an agent talks to another agent: it hands it a whole task, and the other one decides how to solve it, with its own tools and its own memory.