AI agents are becoming increasingly useful around business applications, but building one for an ERP system involves much more than connecting an LLM and sending prompts.
There is the ERP integration itself, understanding business context, working with DocTypes, retrieving live data, tool execution, and deciding how an agent should interact with actual ERP operations.
We have been working on this problem around ERPNext, and decided to make the core AI Agent project free and open source so other Frappe/ERPNext developers can explore the implementation and build on top of it.
The repository includes a working foundation for:
-
Connecting an AI agent with ERPNext
-
Working with ERPNext data and APIs
-
Agent tools and function execution
-
Business-context handling
-
Natural-language interaction with ERP data
-
Extending the agent with your own ERPNext capabilities
Rather than treating it as a finished black-box product, the idea is to provide source code developers can actually read, run, modify and extend for their own ERPNext use cases.
The project is Noviz AI, and the open-source code is available here:
GitHub: Git Open source
Contributions, forks and technical feedback from the Frappe community are welcome.