Frappe_Assistant_Core

Frappe Assistant Core exposes ERPNext functionality to any MCP-compatible language model. It acts as an integration layer that registers ERPNext operations as MCP tools (for example, document creation, updates, searches, reporting, and analytics). The server is designed to work with LLMs, Claude Desktop, and other MCP-enabled agents, enabling a secure, auditable interaction with your ERPNext data. Tools are organized in a registry so your LLM can request a specific operation, and the server enforces ERPNext permissions and audit logging to maintain compliance. You can also leverage the one-click Claude setup to generate DXT files for Claude Desktop integration, enabling seamless local or desktop-based LLM workflows. In practice, you can ask the LLM to create documents, run reports, execute Python-based data transformations, or visualize results, and the MCP layer will validate permissions and perform the actions against ERPNext.

To use it, connect your MCP-enabled model or Claude Desktop to the MCP Server endpoint. The LLM issues commands like create_document, search_documents, run_report, or python_execute, and the Tool Registry routes these requests to the appropriate Core Plugins or Plugins. The architecture ensures that every operation is checked for authentication, role-based access, and ERPNext permissions, with all actions logged in the audit trail for traceability.

Prerequisites

• A ready ERPNext/Frappe environment (bench and a site) up and running
• Python 3.8+ and Node.js as required by your ERPNext version
• Access to install apps on your site (bench) and internet access to fetch the Frappe Assistant Core app

Step-by-step installation

1. Get the application into your ERPNext bench environment

• Ensure you are in your frappe-bench directory

# 1. Get the app
cd frappe-bench
bench get-app https://github.com/buildswithpaul/Frappe_Assistant_Core

2. Install the app on your site

• Replace [site-name] with your actual site name

# 2. Install on your site
bench --site [site-name] install-app frappe_assistant_core

3. Configure and run

• Ensure your site is running and accessible
• Start any required services for your environment and verify that the MCP endpoint is reachable
• If needed, set up DXT/Claude integration via the provided tooling in the app

4. Optional: Run via MCP-compatible command

# Example from the MCP config (adjust as needed for your environment)
pip install -r requirements.txt
python -m frappe_assistant_core

Prerequisites recap

• Installed ERPNext via bench (bench, frappe, Redis, MariaDB, etc.)
• Access to install Python packages and deploy apps
• Network access to fetch the Frappe_Assistant_Core repository

Notes and tips:

• Environment variables: set FRAPPE_SITE and FRAPPE_SITE_BASE_URL to point to your site, and MCD_LOG_LEVEL for debugging verbosity.
• Access control: the server enforces ERPNext permissions and an audit trail. Ensure the user/role you use has appropriate permissions for the tools you intend to expose.
• Claude Desktop integration: use the One-Click DXT setup to generate Claude Desktop integration files for quick, local LLM usage.
• If you run into authentication issues, verify that the OAuth2/OpenID Connect configuration is correctly wired and that Dynamic Client Registration endpoints are reachable.
• For development, you can enable verbose logging by adjusting MCP_LOG_LEVEL to DEBUG and tailing the logs to diagnose tool permission or ERPNext permission issues.
• Tool availability: 21+ built-in tools cover document operations, search, reporting, analytics, Python execution, and visualization. Plugins can extend capabilities as your business needs evolve.