Manufacturing ERP, MES, and QMS platform

Contents

• Why Carbon
• Features
• Architecture
• Tech Stack
• Monorepo
• Getting Started
• Local Dev CLI (crbn)
• Environment Variables
• Logging In
• Commands
• API
• Migration Notes

Why Carbon

We built Carbon after years of building end-to-end manufacturing systems with off-the-shelf solutions. We realized that:

• Modern, API-first tooling didn't exist
• Vendor lock-in bordered on extortion
• There is no "perfect ERP" because each company is unique

We built Carbon to solve these problems ☝️

Features

ERP	Sales, purchasing, inventory, items, accounting
MES	Shop floor execution and job operations
QMS	Inspections, non-conformances, CAPAs
MRP	Material requirements planning
Traceability	Full lot and serial tracking
Nested BoM	Multi-level bills of material
Configurator	Product configuration
Capacity Planning	Scheduling against real resource capacity
Custom Fields	Extend any record
API & Webhooks	Build your own apps on top of Carbon
MCP Client/Server	AI-native integration surface
Accounting	GL, journals, and third-party sync

See the full roadmap for what's next (up next: Simulation).

Technical highlights

• Unified auth and permissions across apps
• Full-stack type safety (Database → UI)
• Realtime database subscriptions
• Attribute-based access control (ABAC)
• Role-based access control (Customer, Supplier, Employee)
• Row-level security (RLS)
• Composable user groups
• Dependency graph for operations
• Third-party integrations

Architecture

Carbon is designed to make it easy for you to extend the platform by building your own apps through our API. We provide some examples to get you started in the examples folder.

Tech Stack

Layer	Technology
Framework	React Router
Language	TypeScript
Styling	Tailwind
Behavior	Radix UI
Database	Supabase (Postgres + RLS)
Auth	Supabase
Cache	Redis
Jobs	Inngest
Email	Resend
i18n	Lingui
Hosting	Vercel
Billing	Stripe
Geometry	Rust (FCL collision + OpenCASCADE CAD)

Monorepo

The monorepo follows the Turborepo convention of grouping packages into two folders:

carbon
├── apps         # applications
└── packages     # shared code

/apps

App	Description	How to run
erp	ERP application	pnpm dev (boots stack + ERP via crbn up picker)
mes	MES — shop floor	pnpm dev (select MES in picker, or both)
academy	Training	pnpm dev:academy
starter	Example app built on the API	pnpm dev:starter
assembler	Geometry service (Rust): STEP → GLB + assembly motion planning	spawned by crbn up (needs a release binary — see Optional: assembler)

pnpm dev runs the per-worktree dev CLI (crbn up). ERP and MES are first-class — the CLI boots the docker stack, applies migrations, regenerates types/swagger, and spawns the selected apps behind portless. The assembler geometry service is spawned too when its release binary is present. Academy and starter are standalone Turborepo entries.

/packages

Package	Description
@carbon/database	Database schema, migrations and types
@carbon/documents	Transactional PDFs and email templates
@carbon/ee	Integration definitions and configurations
@carbon/config	Shared configuration (vitest, tsconfig, tailwind) across apps and packages
@carbon/jobs	Background jobs and workers
@carbon/logger	Shared logger used across apps
@carbon/react	Shared web-based UI components
@carbon/kv	Redis cache client
@carbon/lib	Third-party client libraries (slack, resend)
@carbon/stripe	Stripe integration
@carbon/utils	Shared utility functions used across apps and packages

Getting Started

Prerequisites

• Docker — the monorepo uses Docker for local development
• Node.js v22 (via nvm)
• pnpm (via Corepack — see below; never npm)

You'll also want accounts with the following external services:

Service	Purpose
Posthog	Product analytics platform
Stripe	Payments service
Resend	Email service

Posthog has a free tier which should be plenty to support local development. If you're self hosting and you don't want to use Posthog, it's pretty easy to remove the analytics.

Clone

Clone the repo into a public GitHub repository (or fork https://github.com/crbnos/carbon/fork). If you want to make the repo private, you should acquire a commercial license to comply with the AGPL license.

git clone https://github.com/crbnos/carbon.git
cd carbon

Install

This repo uses pnpm as its package manager. Enable Corepack so the correct pnpm version (pinned via packageManager in package.json) is used automatically:

corepack enable    # one-time: activates pnpm shim from packageManager field
nvm use            # use node v22
pnpm install       # install dependencies

The dev stack (Postgres, GoTrue, Kong, Storage, Inngest, Inbucket, Studio, Realtime) is booted later by crbn up — see Local Dev CLI below. There is no separate "start the database" step.

Optional: the assembler geometry service

assembler is a Rust service (STEP → GLB + assembly motion planning) over C++ FCL and OpenCASCADE. ERP/MES run fine without it — set it up only if you need the 3D /convert and /plan endpoints.

1. Toolchain + native build deps (macOS):

   curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh   # Rust, if not already installed
   brew install fcl cmake ninja draco                               # collision libs (+ libccd/eigen/octomap), build tools, Draco mesh compression
   

   On Linux, install the equivalents from your package manager: libfcl-dev libccd-dev libeigen3-dev liboctomap-dev libdraco-dev cmake ninja-build plus a C/C++ toolchain.

   ./setup.sh already installs Draco on macOS. If yours lives outside the Homebrew keg (/opt/homebrew/opt/draco on arm64), point draco-bridge's build at it with DRACO_PREFIX=/path/to/draco cargo build.

2. Build OCCT once — a patched static OpenCASCADE, cached in ~/.cache/carbon-occt. Slow (~15–30 min) but one-time per machine; re-running is a no-op once cached:

   ./apps/assembler/scripts/build-occt.sh
   

3. Build the service — seconds once OCCT is cached (build.rs finds it automatically):

   cargo build --release -p assembler
   

crbn up spawns the binary when it's present. Verify it's up with curl -sf "$ASSEMBLER_SERVICE_URL/health" (the URL is in your worktree's .env.local) or by watching the asm | lines in the crbn up output. Without the binary the rest of the stack still runs — only /convert and /plan are unavailable.

Local Dev CLI (crbn)

crbn is a small CLI at packages/dev/bin/crbn that wraps two things:

• Git worktrees — every feature branch can live in its own checkout dir, so you can switch branches without stashing.
• Per-worktree docker compose stack — each worktree gets its own Postgres / Supabase services on dynamic ports, isolated under a carbon-<slug> compose project. Routing is handled by portless (a local HTTPS reverse proxy that serves *.dev hostnames on :443 with locally-trusted certs — installed automatically on first crbn up).

Windows users: the dev CLI (crbn, setup.sh) is POSIX-only and expects WSL or Git Bash. Native cmd.exe / PowerShell shells are not supported. From a WSL/Git Bash prompt, the standard flow (./setup.sh, pnpm dev, crbn checkout …) works the same as on macOS/Linux.

Run setup.sh once to put crbn on your $PATH and install the crbn shell function (so crbn checkout can change cwd):

./setup.sh                   # writes a sentinel block to ~/.zshrc or ~/.bashrc
source ~/.zshrc              # or open a new shell
crbn                         # shows commands

Common flows:

crbn checkout sid/cool-thing       # cd into worktree (creates if missing,
                                   # auto-fetches from origin if needed)
crbn checkout -b feat/new-thing    # new branch off origin/main + worktree
crbn checkout sid/cool-thing --up  # …and boot the stack inside it
crbn checkout 760                  # fetch GitHub PR #760 into a `pr-760`
                                   # branch + worktree (fork PRs work too)
crbn copy                          # re-sync .env from main checkout
crbn up | down | reset | status    # per-worktree compose stack
crbn new | list | remove           # interactive worktree management

crbn up flags:

• --no-migrate — skip supabase migration up (use when schema is already current and you just want to re-boot containers fast)
• --no-regen — skip regenerating packages/database/src/types.ts + swagger-docs-schema.ts (auto-skipped when --no-migrate is set, since no schema change implies no type drift)

Files synced by crbn copy are listed under package.json#crbn.copy (defaults to [".env"]). To uninstall the rc block: ./setup.sh --uninstall.

Environment Variables

Create an .env file and copy the contents of .env.example into it:

cp ./.env.example ./.env

Then configure each service:

1. Social Sign In

Signing in requires you to set up one of two methods:

• Email requires a Resend API key (you'll set this up later on)
• Sign-in with Google requires a Google auth client with these variables. See the Supabase docs for instructions on how to set this up:
  • Set Authorized JavaScript origins to https://api.carbon.dev
  • Set Authorized redirect URIs to https://api.carbon.dev/auth/v1/callback
  • About the two API URLs you'll see: each worktree has its own scoped Supabase URL (https://<worktree>.api.dev) for app traffic, and there is one stable alias https://api.carbon.dev registered on whichever worktree is currently up. The stable alias exists only so OAuth callbacks have a single registered redirect URI — one Google Console entry covers every worktree. Day-to-day, your app talks to its worktree-scoped URL; only the OAuth callback hits the stable alias.
• You should set environment variables like the following:
  • SUPABASE_AUTH_EXTERNAL_GOOGLE_CLIENT_ID="******.apps.googleusercontent.com"
  • SUPABASE_AUTH_EXTERNAL_GOOGLE_CLIENT_SECRET="GOCSPX-****************"

2. Supabase

Backend services run inside the per-worktree docker stack — crbn up boots them and writes everything you need into .env.local automatically:

• SUPABASE_URL — portless alias (e.g. https://local-dev.api.dev)
• SUPABASE_ANON_KEY, SUPABASE_SERVICE_ROLE_KEY — keys minted per-worktree from a random SUPABASE_JWT_SECRET
• SUPABASE_DB_URL — direct Postgres URL on a dynamic port

.env.local is generated; do not commit it or hand-edit values that came from crbn up (they are re-derived on each boot). Put genuine secrets (OAuth client IDs, Stripe keys, Resend) in .env only.

Run crbn status at any time to see the live port assignment and the URLs portless is serving.

3. Redis (caching)

No setup needed for local dev — crbn up boots a shared Redis container and writes REDIS_URL into .env.local automatically (each worktree gets its own logical Redis DB). For self-hosted production, set REDIS_URL to any Redis-compatible endpoint (Upstash, AWS ElastiCache, etc.) in your prod environment.

4. Posthog (analytics)

In Posthog go to https://[region].posthog.com/project/[project-id]/settings/project-details to find your Project ID and Project API key:

• POSTHOG_API_HOST=[https://[region].posthog.com]
• POSTHOG_PROJECT_PUBLIC_KEY=[Project API Key starting 'phc*']

5. Stripe (payments)

Create a Stripe account, add a STRIPE_SECRET_KEY from the Stripe Settings > Developers interface:

• STRIPE_SECRET_KEY="sk_test_*************"

6. Resend (email)

Create a Resend account and configure:

• RESEND_API_KEY="re_**********"
• RESEND_DOMAIN="carbon.ms" (or your domain, no trailing slashes or protocols)
• RESEND_AUDIENCE_ID="*****" (Optional — required for contact management in packages/jobs)

Resend is used for transactional emails (user invitations, email verification, onboarding). All three variables are stored in packages/auth/src/config/env.ts.

Finally, boot the stack and the apps:

pnpm dev                # equivalent to `crbn up` — picker lets you choose ERP/MES

crbn up prints a summary box with the live URLs once the stack is healthy. Defaults look like:

Surface	URL
ERP	https://<worktree>.erp.dev
MES	https://<worktree>.mes.dev
Supabase API	https://<worktree>.api.dev
Supabase Studio	https://<worktree>.studio.dev
Inngest	https://<worktree>.inngest.dev
Mail (Inbucket)	https://<worktree>.mail.dev
Postgres	postgresql://postgres:postgres@localhost:<PORT_DB>/postgres

<worktree> is derived from the branch name (e.g. sid-local-dev → local-dev). The main checkout drops the prefix and just uses erp.dev, mes.dev, etc. Ports for raw TCP services (Postgres, Inbucket, Inngest) are dynamic per-worktree — crbn status is the source of truth.

Academy and starter still run on classic localhost ports via pnpm dev:academy / pnpm dev:starter (they are not part of the per-worktree stack).

Logging In

For local development you don't need email or OAuth configured. crbn up seeds a smoke-test user (test@carbon.ms) and writes DEV_BYPASS_EMAIL=test@carbon.ms into .env.local for you. When that bypass email is set, signing in with it skips the magic link and logs you straight into the ERP:

1. Open the ERP at the URL from the crbn up summary (e.g. https://<worktree>.erp.dev/login).
2. Type test@carbon.ms into the email field.
3. Click Sign in with Email.

You'll land on the authenticated dashboard (/x) — no inbox check required. The same session cookie works for the MES app at https://<worktree>.mes.dev.

The bypass only applies to the exact address in DEV_BYPASS_EMAIL and only when that user is active — it's a dev convenience, not present in production. Any other email falls back to the normal magic-link / verification flow (which needs Resend configured). To sign in as your own account instead, use the magic link and read it from the local mail catcher at https://<worktree>.mail.dev.

Code Formatting

This project uses Biome for code formatting and linting. To set up automatic formatting on save in VS Code:

1. Install the Biome VS Code extension
2. Add the following to your VS Code settings (.vscode/settings.json or global settings):

"editor.codeActionsOnSave": {
  "source.organizeImports.biome": "explicit",
  "source.fixAll.biome": "explicit"
},
"editor.defaultFormatter": "biomejs.biome"

Commands

Command	Description
pnpm dev	Boot the stack + apps (crbn up picker)
pnpm run db:function:new <name>	Add an edge function
pnpm run db:migrate:new <name>	Add a database migration
pnpm run agent:new <name>	Add an AI agent
pnpm run tool:new <name>	Add an AI tool
crbn down	Stop the stack (keeps volumes — data preserved)
crbn reset	Wipe the stack and start clean (destroys Postgres volume + flushes the redis db for this worktree)
pnpm db:types	Regenerate types → packages/database/src/types.ts + functions/lib/types.ts (normally crbn up does this after applying migrations)
pnpm generate:swagger	Regenerate swagger → packages/database/src/swagger-docs-schema.ts
pnpm --filter <pkg> <cmd>	Run a command against a single workspace, e.g. pnpm --filter @carbon/react test

Restoring a production snapshot

To restore a production database snapshot locally, use crbn restore. It handles both plain-text .backup and custom-format .dump archives, drops and rebuilds the public schema, realigns internal sequences, resets storage metadata, then applies any migrations the backup predates and regenerates types.

1. Export a backup from your production Supabase project (pg_dump or Supabase Dashboard → Database → Backups).

2. Run it from your worktree root:

   crbn restore /path/to/db_cluster.backup
   # …or for .dump archives:
   crbn restore /path/to/postgres_YYYYMMDD.dump
   

   It prompts before replacing the database. The stack must already be running (crbn up) — a restore rewrites the auth and storage schemas, which GoTrue and Storage build through their own migrations when those containers boot, so crbn restore refuses rather than restore into an uninitialized stack.

   To also get local admin access, pass your production email — your account is upgraded to Admin in the companies it already belongs to and the password is reset locally:

   crbn restore /path/to/backup.backup --admin-email you@example.com
   # Optional: set a custom local password (default: localpass)
   crbn restore /path/to/backup.backup --admin-e