// for developers

Your editor.
Your terminal.
Your agent.

dForge meets you where you already work. Syntax-highlighted, schema-validated module files in VS Code. A CLI on npm you run with npx or install locally. An MCP server that lets your AI agent drive the whole module surface. And an importer that turns an existing database into a working module.

toolchain
● ready
editor vs code · zed
cli npx · npm i -g
agent mcp · claude code
import sql · dbml
/ 01 · editor

Author modules
where you
already write code.

The dForge VS Code extension understands every file in a module. DSL syntax highlighting, JSON-Schema validation, and inline errors across entities, views, roles, actions and triggers — so a typo is a red squiggle, not a failed install.

Get the VS Code extension →
[highlighting]
DSL syntax highlighting

Entities, fields, formulas and the action DSL are colorized so module files read like real code, not config soup.

[validation]
JSON-Schema validation

Every module file is validated against its schema as you type. Required keys, enum values and field types are checked before you ever package.

[navigation]
Errors inline

Bad references, unknown field types and malformed views surface in the Problems panel with the exact line — caught in the editor, not at install.

[editors]
More than VS Code

Zed highlighting ships today and more editors are on the way. The schemas are open, so any JSON-Schema-aware editor gets validation for free.

crm.module / entities / lead.json 0 problems
{
  "entity": "lead",
  "label": "Lead",
  "fields": [
    { "name": "company", "type": "text", "required": true },
    { "name": "stage", "type": "flags", "options": ["new","won"] },
    { "name": "owner", "type": "ref", "to": "user" }
  ]
}
/ 02 · cli

One command
to scaffold,
validate & ship.

The dForge CLI lives on npm. Run it ad-hoc with npx — no install — or add it to a project and wire it into your scripts and CI. Scaffold a module, validate it against the schemas, package a versioned .dforge artifact, and install it into any tenant.

[npx]
Run it with npx

No global install, always the latest version. Best for scaffolding a new module or a one-off validate / package run.

[local]
Install it locally

Add it as a dev dependency and call it from npm scripts and CI. Pin a version so every developer and every pipeline runs the same toolchain.

what the CLI does
  • scaffold a new module workspace with the right layout and schemas
  • validate every file against the dForge JSON-Schemas
  • package a versioned .dforge artifact
  • install or upgrade a module into a tenant
~/acme · dforge cli exit 0
$npx @dforge/create-module acme
✓ scaffolded acme.module · schemas wired
$npm run validate
✓ 14 entities · 9 views · 3 roles · 0 errors
$npm run package
✓ acme@1.0.0 → acme.dforge
/ 03 · ai-native

Let your agent
drive the
whole surface.

The dForge MCP server exposes the entire module surface to AI agents like Claude Code. The agent adds entities and fields, builds views, wires roles, writes actions and triggers, then packs and installs — calling real tools, not pasting guesses. The built-in skill knows every convention.

[entities]
Model the data

Create entities and fields, add references and dependencies — the agent calls typed tools, so the metadata is always valid.

[surface]
Build the surface

Views, folders, menus, roles, actions, triggers, reports and webhooks — the whole module is addressable from the agent.

[lifecycle]
Pack & install

Inspect, pack a versioned artifact, and install it into a tenant — the build-and-ship loop runs end to end from the conversation.

claude code · dforge-mcp skill loaded
you

Add an SLA timer field to the ticket entity and a kanban view by status

claude
✓ entity_field_add: ticket.sla_due (datetime)
✓ view_add: ticket_board (kanban by status)
✓ module valid · 0 errors
✓ packed ticket@1.4.0 → install ready
/ 04 · import

Already have
a database?
Start from it.

Point the schema importer at an existing SQL or DBML schema and get a rough-cut module — entities, views, menus, folders and roles inferred for you, your legacy data carried along. Let AI finish it, validate, and ship. You don't start from a blank file.

from schema to module
  • 01 feed it a SQL dump or a DBML file
  • 02 entities, views, roles and folders are inferred
  • 03 AI finishes the rough cut in your editor or agent
  • 04 validate, package, install
~/acme · import ✓ module ready
$npx @dforge-core/schema-importer acme --from-dbml schema.dbml
✓ 14 entities · 9 views · 3 roles → acme.module
/ 05 · build something

Pick a tool.
Start building.

Scaffold a module with the CLI, open it in your editor, or hand the whole thing to your agent. Read the docs or talk to us about a walkthrough with your team.