buildswithpaul/Frappe_Assistant_Core

Infrastructure that connects LLMs to ERPNext. Frappe Assistant Core works with the Model Context Protocol (MCP) to expose ERPNext functionality to any compatible Language Model

What it solves

Frappe Assistant Core (FAC) allows users to interact with their ERPNext site using natural language. It bridges the gap between LLMs (like Claude or ChatGPT) and ERPNext data, enabling users to perform complex tasks—such as retrieving overdue invoices, updating lead statuses, or running revenue reports—without manually navigating the interface.

How it works

FAC acts as a gateway that exposes ERPNext data and functionality via the Model Context Protocol (MCP). It provides 24 built-in tools for document CRUD, search, reports, and analytics.

There are two primary ways to use it:

  1. BYO-LLM (MCP server): Users connect an MCP-ready LLM client (e.g., Claude Desktop, Cursor) to their site. The LLM uses the user's own API keys and the data remains within the user's stack.
  2. FAC Chat (SaaS): An opt-in, managed experience that provides a native chat widget and a full-screen SPA inside the Frappe Desk, powered by FAC Cloud.

Authentication is handled via OAuth 2.0, ensuring the LLM only accesses data the specific user is permitted to see based on their ERPNext roles. All actions are recorded in an Assistant Audit Log.

Who it’s for

Administrators and users of ERPNext/Frappe sites who want to integrate AI assistants to automate business workflows and data retrieval.

Highlights

  • MCP Integration: Native support for the Model Context Protocol for seamless connection to external LLM clients.
  • Live Schema Access: Reads metadata directly from Frappe's API, meaning custom fields and DocTypes are immediately visible to the LLM without needing a sync.
  • Extensible Toolset: Developers can add their own tools and skills via hooks in their own Frappe apps.
  • Permission-Aware: Strictly adheres to ERPNext user permissions via OAuth 2.0.
  • Built-in Analytics: Includes tools for running Python code and database queries for advanced data analysis.

Related

  • Project
    cbcoutinho/nextcloud-mcp-server

    A production-ready MCP server that connects AI assistants to Nextcloud, enabling LLMs to manage notes, calendars, contacts, and files via natural language.

  • Project
    salesforcecli/mcp

    An MCP server that enables LLMs and AI agents to securely read, manage, and operate Salesforce orgs through a standardized set of tools.

  • Project
    ivnvxd/mcp-server-odoo

    An MCP server that enables AI assistants to interact with Odoo ERP systems, allowing users to search, create, and manage business data via natural language.

  • Project
    marianfoo/sap-ai-mcp-servers

    A curated directory of open-source MCP servers, AI skills, and plugins that enable AI agents to interact with and automate SAP software and development workflows.

  • Project
    freee/freee-mcp

    A bridge that lets AI agents (Claude, Codex, etc.) call freee’s business‑software APIs via a secure MCP server and a set of “Agent Skills” describing each endpoint.