Open Source · AI-Native CRM

Learn to build
real software.
Start here.

A full-stack AI-Native CRM you can take apart, modify, and extend. 236 files. 24,000 lines. Not a tutorial — a working enterprise application built the way real products are built.

PowerShell
❯ git clone https://github.com/DT-CultureTech/DT-CRM-Opensource.git
❯ cd DT-CRM-Opensource
❯ npm install
❯ npm run dev
✓ Server running on :3001
✓ Client ready at localhost:5173
The Origin Story

One person. One AI session.
85 minutes. 236 files.

On June 7, 2026 at 8 AM, we started with one sentence: "I want a sandbox for people to learn web dev." No wireframes. No Jira. No spec. Just an intent and an AI pair programmer. Here's what happened.

8:00 AM
Intent

"I want a sandbox for people to learn web dev."

8:15 AM
Debated ERP vs CRM

CRM won — more learning per unit of domain complexity.

8:25 AM
Defined AI-Native

Not AI-Added. Remove the AI and the product degrades.

8:35 AM
Designed the ontology

Two separate systems — backend reasons, frontend perceives.

8:50 AM
Four parallel agents

Schema, types, server (79 files), client (95 files).

9:25 AM
236 files pushed

24,189 lines. Working AI-Native CRM. Done.

The 35 minutes of conversation — debating, defining, designing — produced the clarity that made 50 minutes of parallel code generation possible. The thinking was the hard part. The code followed.

The Stack

Real tools. Real patterns.

The same stack used at companies from seed stage to IPO. Nothing dumbed down, nothing simulated.

React + Vite
User Interface

Every modern web app is built with this — component-based, fast hot-reload.

Node.js + Express
Backend Server

Handles business logic, API routes, authentication, and AI orchestration.

PostgreSQL + Prisma
Database + ORM

Where your data lives permanently. Prisma lets you talk to it without writing SQL.

Redis
Speed Layer

Caching, real-time pub/sub, background job queues, rate limiting.

WebSockets
Real-Time Updates

Data changes appear instantly — no refresh needed.

AI Integration
Intelligence Layer

Three modes: mock (free), local (Ollama), cloud (OpenAI/Anthropic).

Architecture

This isn't another CRUD app.
This system thinks.

Backend — "What does the system know, and how does it reason?"
substrate/
Infrastructure

Database, cache, AI provider, auth, events, WebSocket — the pipes every domain uses.

domains/
Business Logic

Contacts, Deals, Activities — each with its own intelligence/ surface. AI is distributed, not centralized.

reactions/
System Thinking

Event-driven side effects. "Activity logged" → re-score deal, analyze conversation, generate suggestions.

suggestions/
Human-AI Contract

Every AI inference needing human approval flows here. The system proposes, you decide.

Frontend — Every component is classified by its interaction mode
human/
Your Input

Forms, filters, actions — what you type into the system.

display/
Confirmed Facts

Cards, timelines, charts — data the system knows is true.

intel/
AI Suggestions

Enrichment, scoring, drafts — with approve/dismiss on each.

The Litmus Test

Delete every intel/ folder. The app still compiles and runs — a clean traditional CRM. Add them back and the system thinks. Intelligence is woven through but cleanly separable. Build the CRM first, then add AI surface by surface.

Setup

From zero to running in 15 minutes

01

Install the tools

  • Node.js (nodejs.org)
  • VS Code (code.visualstudio.com)
  • Git (git-scm.com)
  • PostgreSQL & Redis
02

Clone and configure

  • git clone the repo
  • npm install
  • Copy .env.example → .env
  • Set your database password
03

Set up the database

  • npx prisma migrate dev
  • npm run db:seed
  • 4 demo users, 20 contacts, 12 deals, 40 activities
04

Start building

  • npm run dev
  • Open localhost:5173
  • Log in as admin@demo.com / demo123
  • You're inside an AI-Native CRM
Copy and paste into PowerShell
❯ git clone https://github.com/DT-CultureTech/DT-CRM-Opensource.git
❯ cd DT-CRM-Opensource
❯ npm install
❯ Copy-Item .env.example .env
❯ # Edit .env — set your DATABASE_URL with your PostgreSQL password
❯ npx prisma migrate dev --name init
❯ npm run db:seed
❯ npm run dev
Learning Path

Four weeks to thinking
like a builder.

Week 1

Observe

Use the CRM. Create contacts, move deals, log activities. Watch the AI enrichment happen. Don't touch the code yet — just notice what the system does automatically.

Week 2

Read

Open shared/types/ — the nouns. Then server/reactions/ — the verbs. Trace one feature from button click to database write to AI inference to WebSocket push.

Week 3

Modify

Change a deal stage name. Add a field to the contact form. Edit a prompt template. Small changes that teach you how the code connects to the screen.

Week 4

Build

Add a feature: a new dashboard report, a notification when a deal goes cold, a weekly AI summary. Full vertical slice — database to UI.

Who This Is For

Two kinds of builders

Founders & Domain Experts

You have 20 years of domain expertise.
Your tech team has 2 years of context.

The cost of writing code is collapsing. The value of knowing what to build is exploding. When you can read and modify code — even at a basic level — you eliminate the translation layers between your insight and what gets built.

This sandbox lets you practice that bridge. You don't need to become an engineer. You need to become literate enough to close the gap between domain knowledge and product.

Read the full guide
Students & Intrapreneurs

Build at the intersection of
tech, psychology, and business.

The best software is a deep understanding of human behavior, encoded into a system. A CRM is deceptively rich — deal psychology, decision design, trust calibration, behavioral nudges, revenue mechanics. Every feature is a lesson.

This isn't a tutorial that teaches syntax. It's a working enterprise app that teaches you to think like a product builder — the skill stack that compounds: technical execution, system design, AI integration, behavioral design, business modeling.

Read the full guide

Common Questions

The code is open.
Start building.

236 files, a working AI-Native CRM, and an AI pair programmer in your terminal. Everything you need to go from zero to builder.