bench get-app https://github.com/Zeetech-soft-solution/erp-ai-agent

[![Listed on Frappe Gems](https://frappegems.com/api/method/frappe_gems.seo.badge?app=Zeetech-soft-solution%2Ferp-ai-agent)](https://frappegems.com/gems/apps/Zeetech-soft-solution/erp-ai-agent)

ERP Agent AI (Free / Open Core)

An extensible, role-based AI agent for ERP AI integration systems. This is the free, open-source tier: the complete core engine (module registry, entity/workflow/report factories, auth, context assembly, LLM tool-calling loop) plus a working reference connector for ERPNext, with a real, fully working CRM module wired end to end for the Sales User role — leads, contacts, opportunities, customers, and territories, a real lead-qualification workflow, business-rule validation (required contact info, duplicate-record warnings, manager-gated conversion), and the analytics tools (exact sums/counts/averages/correlations, never model-estimated) — plus the folder shape for every other standard ERP module (Selling, Buying, Stock, Accounting, HR, Manufacturing, Projects) present and ready to extend the same way. Sample data and prompts: sample-data/, docs/SAMPLE_PROMPTS.md.

[!TIP] Looking for the full-depth product? erp-agent-pro is the paid tier, built on this exact same open core foundation — no separate fork, only extended capabilities. It takes standard ERP modules beyond basic field mappings, adding deeper business-process logic, domain-specific rules, workflows, validations, and operational capabilities. It provides an extensive MCP tool layer across supported modules and includes a dedicated training-data pipeline for each functional domain.

The ERPNext connector in this repository is the reference implementation of the integration pattern used by Pro to connect ERP-specific business capabilities to any ERP instance — ERPNext today, other ERP platforms tomorrow — without modifying the underlying core.

🔗 noviz.in — product info · 🚀 Live demo of Pro — sign in with a demo account and try it against a real, running ERP.

Screenshots

Real output from the live Pro demo (Accounts Manager role, signed in against a real ERPNext instance) — this is the same tool-calling/reporting architecture this repo's core implements, shown here running with Pro's full accounting module coverage rather than this repo's CRM module.

Outstanding invoice aging	Multi-chart dashboard (pie + bar + line)

📝 Write-up: AI Agents in the Enterprise: Building Smarter ERP Dashboards with Purpose-Built Tools — the architecture behind the dashboard screenshot above: fetch/shape/render as three separate, composable tools instead of one model call doing everything.

What's here vs. what's not

Every module folder exists under backend/src/config/modules/ and backend/src/erpnext/entityMaps/, but only one is actually populated: CRM — customer, opportunity, contact, and territory entities (plus lead, hand-written for its own qualify/convert logic), real business rules, and a working lead-qualification workflow. Every other module's entities.ts/rules.ts/training.ts is present but exported empty, on purpose — the folder shape is real and complete, the business content isn't. The point of this repo is the pattern: how core stays 100% ERP-agnostic, and how a new module or a new connector (SAP, another ERP) plugs in without touching existing code (see docs/ARCHITECTURE.md).

A private, paid tier (erp-agent-pro) built on this same core expands every module to full ERPNext doctype/API coverage and adds additional connectors. It is not part of this repository — see the callout above.

License

AGPL-3.0 (see LICENSE). If you run a modified version of this code as a network service, you must make your modified source available to your users under the same license.

Structure

backend/       Node/TypeScript API server (core engine, config, connectors, routes)
frontend/
  admin/       Admin console (Vite + React)
  agent/       End-user chat/agent UI (Vite + React)
docs/          Architecture, install guide, testing guide, sample prompts, training plan
sample-data/   A real, small CRM data snapshot — see sample-data/README.md

Getting started

cd backend && npm install && cp .env.example .env && npm run dev
cd frontend/admin && npm install && cp .env.example .env && npm run dev
cd frontend/agent && npm install && cp .env.example .env && npm run dev

See docs/INSTALL.md for the full step-by-step install (backend, database, admin console, agent app, ERPNext connector — including installing ERPNext itself and backing up/restoring its data), docs/ARCHITECTURE.md for how modules, entities, and connectors fit together, docs/TESTING_GUIDE.md for the condensed version of the same install aimed at developers who just want it running, and docs/SAMPLE_PROMPTS.md for prompts to try once it's up.