GeneXus for Agents is a set of skills and tools (such as gxnext CLI) that enable any agent (capable of handling skills and tools) to operate transparently on GeneXus Knowledge Bases. Its use provides the productivity of these tools to evolve projects and then generate them with enterprise-grade quality and security using GeneXus’s deterministic generators.
GeneXus for Agents allows you to interact conversationally through an agentic CLI (Command Line Interface) available on the market, enabling an LLM (Large Language Model) to have complete knowledge of GeneXus and execute real operations on Knowledge Bases.
The basic idea is that by combining four components—an Agentic CLI, a GeneXus skill (called Nexa), the gxnext CLI (which interacts with the GeneXus MCP server)— an LLM can be provided with both the conceptual knowledge of GeneXus and the operations required to interact with a Knowledge Base.
In this model:
-
The GeneXus skill provides the LLM with GeneXus knowledge, including language concepts, object types, properties, usage patterns, and best practices. With this information, the LLM can understand your requests and 'reason' about how to implement solutions within the GeneXus ecosystem.
-
The Agentic CLI (for example, CODA CLI, Claude Code CLI, Codex CLI, OpenCode CLI, or other CLIs compatible with LLMs or agents) serves as the interface between you and the LLM, allowing you to send natural-language requests and receive responses or results.
-
The gxnext CLI is an executable program that acts as a proxy between the Agentic CLI and the GeneXus MCP server. It requires the GeneXus MCP server to be running (it is its client). It's a thin frontend that starts, translates each command into a call to the GeneXus MCP server, prints the result, and exits.
-
The GeneXus MCP server (Model Context Protocol server) is a long-running process (daemon/service) that exposes concrete operations that can be executed in GeneXus, such as creating objects, modifying properties, analyzing a Knowledge Base, executing processes, or querying metadata. It hosts the BL in-process, keeps it available, and executes the operations called by the gxnext CLI.
This approach enables the creation of AI-assisted development environments in which the model not only answers questions but also automates development tasks and manipulates Knowledge Bases in a controlled manner.
This layer provides comprehensive knowledge and detailed instructions such as:
gxnext CLI is a client of the GeneXus MCP server. The GeneXus MCP server is the engine while gxnext CLI is a client of that engine.
It's a proxy between the Agentic CLI and the GeneXus MCP server. It translates commands into requests to the GeneXus MCP server.
This layer allows the execution of actions such as:
- Creating a Transaction object
- Modifying a Property
- Running a Build
The typical flow is as follows:
- You express your intent in the Agentic CLI.
- The LLM, using the GeneXus skill, translates that intent into GeneXus objects represented as text files.
- Those text files live in the filesystem and can be versioned in Git.
- The gxnext CLI translates terminal commands into requests to the GeneXus MCP server.
- The GeneXus MCP Server executes operations such as object creation, process execution, and build triggering, enabling GeneXus to generate and compile source code for the target model.
- Results and errors are returned to the Agentic CLI for further iteration.
- Windows Native: Use GeneXus Next (without Docker). See the installation guide.
- macOS Native: Use the native IDE, with the backend running on Docker. See the installation guide.
- Browser (Docker): Run GeneXus Next in your browser using a containerized environment.
Since GeneXus Next 2026.01
Quick Start: GeneXus for Agents