GeneXus for Agents enables a new way of working with GeneXus: more conversational, more flexible, and more aligned with how teams are incorporating Generative Artificial Intelligence into their daily workflows.

Through skills, tools, and interfaces compatible with standards such as MCP, GeneXus for Agents allows AI agents to interact with GeneXus Knowledge Bases from terminals, command-line tools (also known as CLIs), or code assistants – always within a controlled framework.

What is new about this product lies in the access and the work experience: users can now coordinate, review, and supervise agent-assisted tasks. The essentials, however, remain intact: the GeneXus engine, its deterministic generators, and the guarantees that allow mission-critical software to be created and evolved with confidence.

In this article, we answer some frequently asked questions about GeneXus for Agents:

1. What is GeneXus for Agents?

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.

2. How does it work?

GeneXus for Agents adds an interaction layer so AI agents can operate on the Knowledge Base in a more natural and structured way. To achieve this, GeneXus objects can be represented in a textual format that agents understand more easily, making it simpler to read, modify, and validate changes before integrating them back into the KB.

This approach enables working with GeneXus from modern environments compatible with MCP, expanding how developers and agents collaborate throughout the software development lifecycle.

3. What learning curve does GeneXus for Agents imply for GeneXus users?

For a GeneXus developer, the learning curve is incremental. It is not about abandoning the IDE or the traditional way of working, but about adding new interfaces to access the Knowledge Base: the IDE, command-line tools, and compatible agents. The main change involves learning to coordinate and supervise agent-assisted work, while keeping GeneXus knowledge as the foundation.

4. Does GeneXus for Agents work with large KBs?

GeneXus for Agents is designed to work on existing Knowledge Bases, even complex ones. In large KBs, what matters is that the agent does not act blindly: it queries the KB status, existing objects, defined attributes, and patterns used in other modules. Performance and depth of analysis may depend on the agent used, the infrastructure, and the organization of the KB.

5. Can GeneXus for Agents be used from VS Code?

Yes, it can be used from VS Code if the agent, extension, or CLI used is correctly configured to interact with GeneXus for Agents. In VS Code, you can also install the GeneXus Next extension (to be published in the VS Code marketplace) to have an editing experience for KB text files within VS Code.

6. Can a Web Panel be created with GeneXus for Agents?

The creation of Web Panel forms is currently under development. To closely follow the status of this and other features in progress, you can consult the official documentation: GeneXus Next – Work in Progress.

7. Can GeneXus for Agents read a document with user stories to generate code?

Yes, as long as the agent used has the ability to interpret the document format. GeneXus for Agents does not read documents on its own: it enables the agent to interact with the GeneXus KB. In practice, an agent such as CODA, Claude Code, or Codex can analyze user stories, convert them into change proposals, and then operate on the KB through GeneXus for Agents.

8. Is it recorded in the operation history which user generates each change?

It depends on the agent and the workflow used. With CODA CLI – the recommended workflow for clients with Glob.AI OS licensing – actions are recorded within the work ecosystem defined for GeneXus, including review, impact, and model consistency.

Additionally, when changes are integrated into the KB and GeneXus Server, the user responsible for approving or publishing those changes is identified. If other agents are used, the level of traceability will depend on their capabilities and configuration.

9. Does GeneXus Next include its own agents, or is a subscription to external LLMs required?

No. GeneXus Next does not include or host its own LLM or coding agent.

GeneXus Next incorporates AI-assisted capabilities and can be integrated with agents. For enterprise scenarios, the recommended agent is Globant CODA, for reasons of security, governance, and traceability.

There is no lock-in with a single model provider: other agents such as OpenAI Codex or Claude Code can also be used, provided they are supported and correctly configured. The need for an external subscription will depend on the agent, the LLM provider, and the commercial scheme chosen by the organization.

10. Does GeneXus Next on Docker (Linux) have any limitations compared to GeneXus 18 on Windows?

GeneXus Next can run on Mac, Windows, and Linux, and official Docker images exist for linux/amd64 and linux/arm64. That said, the comparison with GeneXus 18 on Windows must be analyzed by workflow: generation, drivers, external dependencies, extensions, and technologies used by each KB.

For GeneXus for Agents, the central point is that agents operate on the local checkout and that the resulting changes are compatible with the KB and the GeneXus 18 workflow.

11. From which GeneXus 18 upgrade is it possible to use the MCP?

There is no minimum required upgrade of GeneXus 18 to use the MCP. Although internal tests are performed primarily with the most recent upgrades, there are users who have used it successfully with earlier versions such as U10, U13, and U15.

12. Are there limitations in the generation of objects (data or interfaces)?

GeneXus for Agents does not change the GeneXus engine. The runtime is not modified, the generators are not altered, and the core is not rewritten. What changes is the way of accessing and operating on the Knowledge Base.

Therefore, the possibilities and limits of generation continue to be determined by GeneXus, its generators, the available objects, and the version used. The agent can assist in creating, reviewing, or modifying objects, but final validation remains in the hands of the GeneXus engine.

13. Can Unit Tests be created with GeneXus for Agents?

This is an area that is still under development. The goal is for GeneXus for Agents to be able to manipulate all types of valid objects in GeneXus 18 and GeneXus Next, which requires developing grammars for the textual representation of the different object types, and skills to teach an agentic CLI how to correctly manipulate these files. Work is already underway to complete these requirements. As for testing objects, grammars have already been developed for all object types, except for the Test Suite object. Skills for these objects have not yet been defined.

14. Can GAM be used with GeneXus for Agents?

Yes, GAM is compatible with GeneXus for Agents. In the next version, specific skills will be released that will improve the effectiveness of the generated code, both for initializing and configuring GAM and for managing authentication. Additionally, work is underway on a GAM REST API that will allow implementing a complete GAM backend using GeneXus for Agents, although this functionality does not yet have a confirmed release date.

15. Are transactions modified with GeneXus for Agents available in GeneXus 18 and GeneXus Server?

Yes. In the workflow for GeneXus 18 KBs, the agent works on a local copy obtained from GeneXus Next. Once changes are reviewed and approved, they are republished or synchronized with GeneXus Server.

Then, when working from GeneXus 18, those objects will be available as part of the normal KB workflow, provided that the changes are compatible with GeneXus 18 and pass the corresponding GeneXus engine validations.

For more information about GeneXus for Agents, we invite you to consult the FAQs and official documentation available on GeneXus Wiki.

You may also be interested in reading:

GeneXus and Neuro-Symbolic Architecture

The Problem with Prompt-Based Development

GeneXus in the Era of Agentic Development