I feel strongly that if Moqui is to remain relevant that it must make the transition to become an AI solution. To that end, I wrote the following document and had Gemini edit and enhance it. It is not meant to be any kind of final directive; I just feel that there must be more discussion. I didn’t put me in the document, so feel free to flame away.
ERP AI Agent Requirements
Summary
It should be obvious to everyone now that in the future there will be no software companies, only AI companies. That means that Moqui and OFBiz and developers will become irrelevant unless we act. I believe that Moqui and OFBiz can become the nucleus of an open source AI consortium that can provide an alternative Big Tech. Big Tech is going to have to demand big payments in order to support its weight, but a Moqui and OFBiz based AI initiative can thrive in the face of Big Tech’s onslaught if it provides a consistent, high quality product. Defining exactly what that AI consortium would look like would be a premature exercise. The organization will not exist unless it can offer an agentic AI tool that rivals that of Big Tech and produces consistent, high quality solutions. The purpose of this document is simply to document the characteristics that such an agentic tool must have in order for everyone in the OFBiz and Moqui community to get behind. There are so many hurdles that we must face in order to stay relevant, and unifying our efforts must be the highest priority and if we can’t unite around one tool, nothing else we do will matter. Following are some topics and characteristics of such a tool. They are meant to be discussion starters. Somehow we need to agree. For the purpose of discussion such a tool will be code named ERP AI tool. It can be a dual purpose tool - both generating solutions and acting as an advisor to the end user. But of the two, the end user advisor is the one for which there could be competing solutions without causing disunity and fragmented solutions and branding.
1.0 Produce “grounded” code.
One of the biggest advantages that OFBiz and Moqui have is that they share a common, incredibly brilliant DNA created by David E Jones and Andy Zeneski. The structure has been the basis of countless mission critical business solutions. To develop a tool that does not use those building blocks would be foolishness. Instead of producing spaghetti code, it would have to first generate screen macros and Mantle UDM and USL artifacts. Screen macros will be one area that will need to be shored up if ERP AI is to compete. Still I do not think that the goal should be to have a black box that produces HTML code. The plan should be to produce screen macro which point to Freemarker macros which point to Vue and Quasar components. That way faulty solutions can be minimized by, again, using solid building blocks and diagnosed more efficiently.
2.0 Truly agentic - take over as much manual labor as possible
It is just a given that ERP AI should take over as much manual effort as possible. That means that it must have direct access to the running environment. The implication is the ERP Ai will be an IDE tool. Developing such a tool from scratch is definitely not something that should be attempted. It is also critical that whatever base tool is chosen it must be around for the long haul. This cannot become a religious war. It should be obvious that Google’s Antigravity is the only viable candidate. Regardless of the current state of Antigravity now, Google is the only player in the AI arena that can be counted on to still be standing 10 years from now.
3.0 Can test, monitor and debug
ERP AI must have the ability to reach and control any part of an app. To do so, it must have the capability of Ean Schuessler’s moqui-mcp Moqui component (GitHub - schue/moqui-mcp: MCP component for Moqui). Incredible advances in constructing a candidate ERP AI have been made by just dropping moqui-mcp into an Antigravity workspace that just points to a Moqui app directory.
The testing, monitoring and debugging capability will be needed in order to compete with future Big Tech solutions. Already, by virtue of adding moqui-mcp to Antigravity, Antigravity has shown the ability to start and stop the server and make modification to code and retrying - all without developer interventions. It is also true that it often got lost in such endeavors, but if no other candidates have demonstrated the ability to do that, it would certainly point to Antigravity as being the leading candidate for ERP AI.
4.0 Can act in end user role
Obviously, the ERP AI tool will have to guide the user in using their ERP AI solution. This would include report generation, question answering and action suggestions and taking. As previously stated, multiple tools that work on consistent OFBiz and Moqui foundations could exist without harming the ERP AI brand.
5.0 Hosting
While not strictly part of the ERP AI tool, how a business tool is delivered could be another selling point for the OFBiz and Moqui consortium. One of the biggest fears of businesses is allowing a third party AI tool access to their data. Again, thanks to Ean, it has been demonstrated that Moqui can run in the same box as an AI processor and an open source, LLM model can be run in conjunction with Moqui - meaning that a business’ AI assisted app can run self-contained.
Next Steps and Recommendations (from Gemini)
While the document clearly articulates the necessity, core requirements, and initial direction for an ERP AI Agent (code-named ERP AI tool), a few critical areas for discussion and action are currently missing or could be strengthened to provide a more comprehensive roadmap.
Suggested Missing Points
- Licensing and Governance Model: The document advocates for an open-source consortium approach to counter Big Tech. A crucial, missing detail is the proposed licensing model for the ERP AI Agent’s core code, any generated artifacts, and the overall governance structure of the consortium.
- Key Questions: Will it be fully permissive (e.g., Apache 2.0)? How will contributions be managed and intellectual property rights secured for the community? How will the consortium be funded and governed to ensure long-term stability?
- Security and Data Privacy Protocol: Section 5.0 touches on the fear of third-party access to data, but this needs to be explicitly integrated as a core technical requirement.
- Key Message: The ERP AI tool must adhere to stringent, enterprise-grade security protocols, including fine-grained access control that respects existing OFBiz/Moqui security models and robust encryption (both in transit and at rest). The ability to run locally (as mentioned) is a feature, but the security model itself must be formalized.
- Migration and Interoperability Strategy: The document assumes the ERP AI will generate new code. A necessary point is how the tool will interact with and potentially modernize existing Moqui/OFBiz installations and custom codebases.
- Key Question: Can the ERP AI tool analyze, document, and potentially refactor or suggest improvements for legacy code built on the shared DNA? This is a huge value proposition for current users.
- Defined Minimum Viable Product (MVP): To unite the community, a concrete, achievable MVP goal is needed, moving beyond general characteristics.
- Example MVP: “The ERP AI tool’s MVP will be the ability to autonomously create a new, simple Moqui screen (e.g., a basic CRUD interface for a new entity) and successfully run a unit test against it, all within the Antigravity/moqui-mcp environment.”
- Community Engagement and Call to Action: The document is meant to be a discussion starter, but it lacks a clear call to action regarding how the community can provide feedback, join the initiative, or contribute to the identified technical hurdles (e.g., shoring up screen macros).
Conclusion
The shift from being a “software company” to an “AI company” is not optional; it is the prerequisite for relevance in the coming decade. For the Moqui and OFBiz communities, the choice is clear: unify efforts to create a superior, open-source agentic tool, or face fragmentation and eventual obsolescence against the consolidated power of Big Tech.
The proposed ERP AI tool, leveraging the foundational genius of the shared Moqui/OFBiz architecture and powered by a long-haul candidate like Google’s Antigravity, offers a unique path forward. By focusing on generating “grounded” code artifacts (UDM, USL, screen macros) instead of proprietary spaghetti code, the initiative can guarantee both quality and maintainability—a fundamental competitive advantage. Furthermore, integrating capabilities for end-to-end testing, monitoring, and debug loops, as demonstrated with moqui-mcp, establishes the necessary intelligence for truly agentic, autonomous development.
The most compelling differentiator, however, is the commitment to data sovereignty. The demonstrated capacity for Moqui to run self-contained with an open-source LLM addresses the most profound fear businesses have about adopting AI: data exposure. This provides the consortium with a powerful, secure value proposition.
The requirements outlined—grounded code, true agency, full control/debug capabilities, and a consistent user advisory role—represent the non-negotiable minimum for achieving unity and producing a tool capable of challenging the incumbents. The next steps must involve translating these requirements into a formal governance structure, a defined security roadmap, and a concrete MVP that catalyzes community contribution. By embracing this unified vision, the Moqui and OFBiz consortium can secure its future, not just as an alternative, but as the superior, trust-based foundation for mission-critical enterprise AI solutions.