GeneXus
for Agents is the layer that enables Artificial Intelligence agents to understand, generate, review, and propose changes to
GeneXus
objects with full Platform context, while the
GeneXus
engine validates those changes and integrates them back into the development workflow.

In
GeneXus
, all business knowledge lives in the Knowledge Base, not in source code. The KB holds the objects, transactions, rules, data, and relationships that define the system.

GeneXus
for Agents gives AI agents the ability to read, reason over, and propose changes to that KB, while the
GeneXus
engine handles what it has always done: generating deterministic, correct, production-ready code.

How Does
GeneXus
for Agents Work?

The workflow with
GeneXus
for Agents can be summarized as follows:

• An AI agent receives an instruction in natural language (for example, “add a vendor management module with tiered approval”).
• The agent accesses the Knowledge Base through the
  GeneXus
  MCP Server, which implements the Model Context Protocol (MCP). This protocol defines a structured, standard way for AI agents to interact with the
  GeneXus
  Knowledge Base. KB objects are represented as .gx files with
  GeneXus
  ‘s own declarative syntax, allowing language models to read them, reason about them, and propose modifications efficiently.
• The agent works with
  GeneXus
  Skills: a set of instructions that teach the agent how to behave within the
  GeneXus
  ecosystem, which conventions to follow, how to name objects, and which patterns to use.
• Changes suggested by the agent are automatically validated by the
  GeneXus
  engine, though the developer can review and approve each modification before it is integrated into the KB.
• Once validated,
  GeneXus
  generates code deterministically for the target technology, just as it always has.

This workflow can be executed from the
GeneXus
IDE or from a Command Line Interface (CLI).

Why Is
GeneXus
for Agents Important for
GeneXus
Developers?

GeneXus
for Agents empowers developers – it does not replace them. The team remains in charge of deciding what gets built, reviewing what the agent proposes, and making architecture decisions. The difference is that they can now delegate repetitive tasks, explorations, UI generation, auxiliary integrations, and reviews to AI, while
GeneXus
ensures the system remains consistent and maintainable.

It is designed to avoid dependency on any particular AI model, allowing the use of any LLM compatible with MCP. This means system control stays in the hands of the team and
GeneXus
– not the AI provider.

For those evaluating whether to adopt
GeneXus
, this is one more reason to commit: they will get the Platform’s traditional guarantees along with the ability to work with AI in a structured, controlled way.

GeneXus
for Agents is available at no additional cost for users with active maintenance who are running
GeneXus
Next or
GeneXus
18. If you are working with an earlier version (
GeneXus
15, 16, or 17) and want to use this functionality, you will need to upgrade to
GeneXus
18. To migrate to a compatible version, we recommend contacting Fast Code Studio by Globant or
GeneXus
Partners.

GeneXus
for Agents: AI speed, with
GeneXus
guarantees

In the upcoming articles in this series, we will dive deeper into each component of
GeneXus
for Agents: what the Knowledge Base is and why it is key, how the
GeneXus
MCP Server works, what
GeneXus
Skills are, and why prompt-only development has structural limitations that go beyond model quality.

We invite the entire
GeneXus
Community to explore and experiment with their solutions using
GeneXus
for Agents.

For more information, visit the official
GeneXus
for Agents documentation.

You may also be interested in reading:

The Campfire Model: Echoes of the past in the age of agent-driven development

What is
GeneXus
for Agents?

GeneXus
in the Era of Agentic Development