The GEAI Proxy is a Python-based component that enables dynamic integration of external Tools into Globant Enterprise AI via the Model Context Protocol (MCP) and Agent-to-Agent (A2A). It acts as a bridge between GeneXus Enterprise AI and one or more MCP- or A2A-compliant Servers.
Once the MCP/A2A Servers are properly configured and connected through the GEAI Proxy, the Tools they expose automatically appear in the AI and Tools Tab of an Agent, ready for use by any Agent without additional setup.
Note: This documentation applies to GEAI Proxy version 0.3.0.
pip install pygeai
The GEAI Proxy supports integration with MCP-compliant Tool Servers, and also with A2A-compliant external Agents.
It requires two types of configuration:
- Tool Servers configuration (supports MCP and A2A Servers)
- Proxy authentication configuration (to connect with Globant Enterprise AI)
This section defines the MCP and A2A Servers that this proxy will link to Globant Enterprise AI. The configuration follows the Claude Desktop standard format and supports multiple servers in a single file.
{
"mcpServers": {
"serverName1": {
"command": "command-to-launch-server",
"args": ["arg1", "arg2", ...]
},
"serverName2": {
"command": "command-to-launch-server",
"args": ["arg1", "arg2", ...]
}
}
}
{
"mcpServers": {
"puppeteer": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-puppeteer"]
}
}
}
{
"mcpServers": {
"filesystem": {
"command": "npx",
"args": [
"-y",
"@modelcontextprotocol/server-filesystem",
"~/mcp-shared-folder"
]
}
}
}
{
"mcpServers": {
"markitdown": {
"uri": "http://localhost:5000/sse"
}
}
}
{
"mcpServers": {
"puppeteer": {
"command": "npx",
"args": "-y", "@modelcontextprotocol/server-puppeteer"
},
"filesystem": {
"command": "npx",
"args": "-y", "@modelcontextprotocol/server-filesystem", "~/mcp-shared-folder"
},
"markitdown": {
"uri": "http://localhost:5000/sse"
},
"custom-server": {
"command": "python",
"args": "path/to/your/custom_mcp_server.py"
}
}
}
This section establishes the connection between the proxy and Globant Enterprise AI and manages user aliases. You can configure it automatically on the first run or manually reconfigure an existing alias.
geai-proxy sample-mcp-config.json --alias myalias
geai-proxy --configure --alias myalias
During interactive setup, the CLI will prompt you for each required setting:
# Configuring GEAI proxy settings...
Generated new proxy ID: 37bae96b-bc99-4110-bb61-b912b28f9e32
-> Insert proxy ID (UUID) (Current: 37bae96b-bc99-4110-bb61-b912b28f9e32, Leave empty to keep):
-> Insert proxy API key:
-> Insert proxy base URL:
-> Insert proxy name:
-> Insert proxy description:
During interactive setup, the following parameters are requested:
- proxy ID (UUID): A unique identifier for this proxy instance. If left empty, an automatically generated UUID will be used.
- proxy API key: The API key used to authenticate the proxy with the Console. Must be an API token from Globant Enterprise AI for a specific Project.
- proxy base URL: The base URL of the Globant Enterprise AI installation to connect to. For example: https://api.saia.ai.
- proxy description: (Optional) Helps identify this proxy instance when using multiple aliases.
To start the proxy server with a specific configuration and alias:
geai-proxy sample-mcp-config.json --alias myalias
| Argument |
Description |
Required |
| config_file |
Path to the MCP servers configuration file (JSON format) |
Yes |
| --alias |
Alias for the proxy settings |
No |
| --configure |
Command to reconfigure proxy authentication |
No |
geai-proxy sample-mcp-config.json --alias beta
Once GEAI Proxy is running and MCP Servers are connected, the Tools they expose will appear under AI and Tools Tab of an Agent in The Lab. Agents can immediately start using these Tools—no manual import or setup is required.