A2A Server

A2A Server turns a NappAI flow into an agent that other agents can use. A2A (Agent2Agent) is an open standard that lets agents on different platforms talk to each other: Google’s agent tools and the protocol’s official kit, among others, support it. With this component your flow publishes a public card with its name, what it does and usage examples, and any compatible agent can send it tasks.

The agent is the whole flow: the incoming message enters through the Chat Input and the answer is what the Chat Output returns. The component is not connected to anything; it only describes the agent.

Guide with video

The A2A section explains step by step, with a video, how to publish an agent and how to connect to others.

When to use it

• Offer your customer service assistant to a partner’s agents, so they can ask about orders or incidents.
• Let your company’s orchestrator agent, built on another platform, delegate billing tasks to a NappAI flow.
• Publish a specialised agent (for example, contract review) for another team to use from its own tools.
• Connect two NappAI flows in different accounts: one publishes with A2A Server and the other calls it with A2A Agent.

Configuration

Field	What it does
Description (required)	What the agent does and when it is worth calling. Other agents read it to decide whether to hand it a task, so be specific.
Agent Name	The name other agents see. If you leave it empty, the flow’s name is used.
Skills	Optional. One skill per line, written as Name: what it does. If you write none, the description is the only skill.

Advanced

Field	What it does
Example Requests	One example request per line, to guide callers.
Tags	Comma-separated keywords that help others find the agent.
Version	The version shown in the agent card. Defaults to 1.0.0.

Publish the agent

1. Build the flow with a Chat Input, the logic you need, a Chat Output and the A2A Server component.
2. Deploy a version: other agents always use the deployed version. See Versions and publishing.
3. Click API Access and open the A2A Server tab. You will see the address of the agent card, the agent URL and examples to connect from other platforms.
4. Choose an API key: whoever calls the agent needs it. See API keys.

Files

If the caller sends files (a PDF, an image…), the flow receives them in the Chat Input, through its Binary output, just like a file uploaded in the chat. They can come inside the message or as a link to a public internet address.

Files the flow generates, for example an image created by an image generation component, are sent back with the answer. Each file can be up to 20 MB, with at most 5 per message.

Each conversation has its own memory

If the caller continues a conversation, NappAI keeps it: the flow’s memory components remember what was said earlier in that same conversation.

The key gives access to the agent

The agent card is public, but sending it tasks requires an API key of your account. Share it only with whoever should use the agent.

Output

• Agent Card: the card other agents will see, so you can review it in the editor. You don’t need to connect it to anything.

Example

The distributor Ferretería Central has a NappAI flow that checks order status in its ERP. It adds A2A Server with the description “Checks the status and delivery date of Ferretería Central orders from the order number” and deploys. Its main customer, which manages purchasing with its own agent, adds the agent URL and the API key Ferretería Central gave it. From then on, its agent asks about pending orders by itself and gets the flow’s answer.

Related components

• A2A Agent: the opposite direction, so your flows can use agents on other platforms.
• MCPServer: publishes individual tools for assistants such as Claude or Cursor.