Skip to main content

START HERE — the whole platform in one page.

Tu Empresa Digital OS

Tu Empresa Digital OS (TED OS) is a per-client operations brain: collect each client's data, accrete their memory from day one, and run AI tasks over it — then reuse the same core for the next client so each delivery is cheaper than the last. "Service outside, product inside."

This site documents the whole platform — its apps, its packages, its data architecture, and the workflow (claude-kit) that builds it. Everything here describes what the system IS today, derived from the canonical reference docs in the repo's knowledge/ directory.

Internal documentation

This site is access-gated (Clerk). It documents internal architecture in full — including decisions, auth, and deploy topology. No page ever contains real secrets, tokens, keys, or env values — environment variables are referenced by name only.

The shape of the system

LayerWhat it isRead
Platform appsThe operator admin, the superadmin console, white-label client instances, and the API behind themPlatform apps
ArchitectureThe Plan Engine graph (dev-workflow SoT), the product System of Record, per-client data & memory, and the last-mile engineArchitecture
Auth & deployClerk authentication across subdomains, and the Vercel + Fly deploy topologyAuth & deploy
Design systemThe neutral foundation, per-surface skins, and the Grid System v2Design system
Build & workflowThe effort model and the kit ↔ engine boundaryBuild & workflow
claude-kitThe AI dev team — agents, skills, rules, and a workflow inside Claude Code — that builds TED OSclaude-kit

Two databases that never mix

A foundational rule runs through the whole platform: the dev-workflow data and the client/product data live in separate databases and never share tables.

  • The Plan Engine graph (plan schema) is the source of truth for our plans and tasks. It doubles as a training-trace corpus for a future builder model.
  • The product System of Record (product schema) holds clients' data, isolated per client.

Tenant data never enters the dev-workflow corpus. See Plan Engine and Product System of Record.

The first client

Comprender (an education provider) is tenant #1 — the case study that drove the product System of Record, per-client isolation, and the white-label client app pattern. See Comprender.

How to read these docs

  1. This Overview for the shape of the system.
  2. Architecture for how it is built.
  3. Platform apps for what each surface does.
  4. Decisions (ADRs) when you need why a thing is the way it is.

The repo's knowledge/ directory is the upstream source of truth; these pages link and derive from it rather than duplicating it.