This article describes how to integrate an external Agent that supports the A2A protocol into your Globant Enterprise AI Project.
From the perspective of Globant Enterprise AI, an external Agent integrated through A2A behaves like a Tool and can be invoked by Agents to resolve tasks.
To integrate an Agent into your project, follow the steps below:
- Get the Agent Card of the external Agent.
- Register and invoke the external Agent in the GEAI Proxy configuration.
To start the integration, you first need to obtain the Agent Card of the external Agent you want to integrate.
The Agent Card is a public JSON document that describes the Agent’s identity, available skills, supported features, and how to authenticate requests. It follows the A2A standard format.
This file is usually located at the following URL:
<AGENT_BASE_URL>/.well-known/<agent_file>.json
Send an HTTP GET request to the Agent Card URL.
If the request is successful, you will receive a JSON response that looks like this:
{
"name": "external_agent",
"description": "A sample external Agent",
"url": "http://external-agent-domain.com",
"version": "1",
"skills":
{
"id": "translate_text",
"name": "Translate Text",
"description": "Translates text from one language to another",
"tags": ["language", "translation"
}
],
"capabilities": {
"streaming": false
},
"securitySchemes": {
"api_key": {
"type": "apiKey",
"name": "Authorization",
"in": "header"
}
},
"defaultInputModes": "text",
"defaultOutputModes": "text"
}
In this case, the name of the Agent in this Agent Card is external_agent and it has a single skill called Translate Text, which translates text from one language to another.
You will use this information in the next step to register the Agent in your GEAI Proxy configuration.
Once you have the Agent Card, the next step is to indicate to Globant Enterprise AI how to access and use that Agent.
To do this, you must register the Agent in the GEAI Proxy configuration.
The GEAI Proxy is a service that connects external systems, such as A2A Agents, with your Globant Enterprise AI Project. It reads a configuration file and registers the declared Agents as Tools in your project.
For Agents that support the A2A protocol, you must declare them under the a2aServers section of the configuration file.
To configure this proxy, edit a JSON file where all the external Agents are declared.
Open your GEAI Proxy configuration file and add an entry for the Agent under the a2aServers section.
Here is a sample configuration:
{
"a2aServers": {
"hello-world": {
"public_prefix": "com.genexus.a2a.sampleagent",
"url": "http://localhost:9999",
"headers": {
"Authorization": "Bearer fh84ff...."
}
}
}
}
Where:
- url: Base URL where the Agent is hosted.
- headers: (Optional) Used to define the authorization or custom headers required to access the Agent.
- public_prefix: (Optional) Used to expose the Agent's skills under a global identifier.
After saving the configuration, start the GEAI Proxy. Run the proxy from the command line as follows:
geai-proxy geai-config.json --alias myalias
The proxy reads this file, connects to the Agent using the given URL and headers, and registers the Agent’s skills in your project.
The Agent’s skills become available in the AI and Tools Tab of an Agent in The Lab.
For example, if the Agent has a skill called translate_text, it will be available in The Lab as a Tool like:
com.genexus.a2a.sampleagent.translate_text
See the official A2A documentation for more details about the Agent invocation protocol.
Since version 2025-07.
How to expose an Agent using the A2A protocol