This is a concept application
⚠️ Work in Progress - This is a concept application exploring the future of visual workflow automation. We're actively seeking feedback, contributions, and criticism to improve the design and implementation.
The AI industry is exiting its "✨ one‑click magic ✨" phase and moving toward reliable, human‑in‑the‑loop tooling.
Developers no longer want opaque agents that vanish into 20‑step misfires or rack up $50 chat sessions—they want intelligent workspaces that:
- Ask clarifying questions
- Surface intermediate results
- Keep them firmly in control
Structured, spec‑driven assistance that lets engineers stay the architect, while AI handles the tedium. Ordenado brings that same philosophy to workflow automation.
✅ Accessible but ❌ crumble on advanced logic (loops, conditionals, data transforms)
✅ Powerful but ❌ slow to wire together, hard to debug, and siloed per developer
✅ Promised autonomy but ❌ proved brittle: error compounding, ballooning token costs, and lack of transparency
Teams need a system that combines visual clarity, code‑level power, and AI help—without the black‑box risks.
| Principle | Implementation |
|---|---|
| Keep humans in the loop | All complex logic lives in Code Nodes you can read, edit, or AI‑generate—but must approve |
| Minimal surface area | Only two node types: Tool Node (calls any installed integration or nested workflow) & Code Node (TypeScript) |
| Show your work | Step‑by‑step execution with cached outputs; click any value to trace its origin |
| Composable by default | A finished workflow can be dragged onto another canvas as a Tool Node—promoting reuse without extra boilerplate |
| Data binding made easy | Inline "chips" reference $PreviousStep.field[0] with live preview; short JSONPath‑lite expressions cover 90% of mapping needs, leaving only edge cases to code |
| AI as copilot, not autopilot | One‑click "Generate code" or "Map these fields" inserts draft snippets; the user confirms or edits. No hidden steps, no unbounded agent loops |
- Drag‑and‑drop canvas powered by React Flow
- Installed‑tool library sourced from your Deco workspace (AI models, databases, HTTP, email, etc.)
- Monaco‑based Code Node with TypeScript type‑checking
- Mock AI templates (swappable for real LLMs)
- Topological runner & debugger with output caching in IndexedDB
- Import/Export JSON DSL v0.2.0 – plain text, version‑controlled, migration‑ready
| Trend | Ordenado's fit |
|---|---|
| Human‑centric AI | Every action is inspectable; AI suggestions are never applied blindly |
| Spec‑driven development | Workflows are themselves specs in JSON; Code Nodes compile those specs into action |
| Micro‑agents & bounded scope | Tool + Code architecture encourages short, verifiable chains, avoiding the error snowball that plagues long autonomous runs |
| Reusable building blocks | Nested workflows act like functions; teams build libraries instead of copy‑pasting zaps |
| Cost transparency | No hidden chat loops; each node run = one deterministic cost |
| Q3 '25 | Q4 '25 | 2026+ |
|---|---|---|
| • Live MCP execution back‑end • Real LLM integration with guardrails • Data‑binding UI (chips + expression preview) |
• Role‑based sharing & comments • Scheduled runs & webhooks • AI "explain my flow" documentation |
• Marketplace for community workflows • Multi‑tenant serverless hosting for micro‑flows • Model‑aware cost optimizer |
Ordenado turns the hype curve into usable software.
It gives developers and power users a single canvas where visual clarity, code precision, and AI assistance coexist—every step transparent, every failure debuggable, every success reusable.
In a market shifting from "autonomous miracles" to trustworthy copilots, Ordenado is the workflow builder that actually works.
- Node.js 18+
- npm or yarn
# Clone the repository
git clone https://github.com/your-username/ordenado.git
cd ordenado
# Install dependencies
npm install
# Start development server
npm run dev# Run in development mode
npm run dev
# Build for production
npm run build
# Deploy to production
npm run deployWe welcome contributions! Please see our Contributing Guidelines for details.
This project is licensed under the MIT License - see the LICENSE file for details.