AI-Native× Execution Loop× Open Source

Your Enterprise
AI Executor

From operations to development—just say the word. OpenOBA gives every enterprise an AI Executor that understands your business and gets things done.

Request Demo GitHub
90%+
3-Year TCO Reduction

Traditional outsourcing vs. OpenOBA—an order of magnitude lower

2,880×
Dev Speed Boost

New field: 2 days → 1 minute · Workflow: 3 days → 30s

100%
Real-World Validation

Eyewear ERP to dogfooding—product is the case, case is the product

300+
Commits, 8 Releases

Built by 1 human + 1 AI Executor. 3 months, full stack.

10 modules eyewear ERP | 300+ commits in 3 months | 8 releases from V1.3 to V1.6 | MIT open source
Product

Not a Chatbot. An Executive.

Most AI tools stop at conversation. OpenOBA goes further—your AI Executor understands your business, operates your systems, and delivers deterministic results. Full audit trail, every step traceable.

Operations

Natural Language Operations

"Add a seasonal field to all eyewear products"—what used to take a dev ticket now takes a single sentence. AI Executor analyzes, modifies code, and verifies results.

Development

AI-Assisted Development

"Add self-manufactured/OEM attribute to all products"—agent handles full cycle: analysis, code changes, commit. From idea to deploy in one conversation.

OpenOBA vs. General AI
General AI Assistant
Conversation only, no system access
No understanding of business context
Prompt-level "constraints"
No audit trail or rollback
OpenOBA
Natural language → Database execution
ERDL deterministic understanding
Action Guard protocol-level validation
Full audit trail, one-click rollback
Core Technology

ERDL · Enterprise Resource Definition Language

In the age of AI, humans write YAML. Digital systems speak ERDL. A cross-party language protocol—human-readable, machine-readable, readable between machines. Transparent communication, full navigation for AI execution, the final piece of enterprise AI adoption.

YAML-Compatible MIT Licensed Hot Reload Entity + Alias + Rule

Industry Alias System

"Frame shape" → shapeCode. "Selling price" → retailPrice. ERDL maps industry jargon to system fields deterministically—no guesswork, no prompt engineering.

Action Guard

Protocol-level validation between LLM output and system execution. Not prompt constraints—hard checks. Every action validated before execution.

Hot Reload

Edit .erdl, save, instant. No restart, no deploy. ERDL Registry watches file changes → validates → hot-swaps. Business rules from "release cycle" to "save-and-go".

ERDL Architecture
ERDL Architecture
Cases

Talks Cheap. Show Me the Code.

Two cases—one external deployment, one internal dogfooding—show what OpenOBA can actually do.

LiveMIT Open Source

Eyewear Industry ERP

A complete ERP system for the eyewear industry—products, orders, customers, inventory. Full digital loop from a single .erdl file.

10 modules
MIT licensed
ERDL-driven

Pain Point

Eyewear SKUs are highly fragmented—frame × color × size × material. Traditional ERP customization is expensive and inflexible.

Solution

A single eyewear.erdl file defines fields, aliases, and rules. OpenOBA auto-generates the complete ERP—SPU/SKU management, color systems, pricing engine.

Result

A fully functional ERP spanning 10 modules. MIT licensed, deploy anywhere, modify freely. Zero-to-production with one industry ERDL file.

Modules
Products Orders Customers Inventory Service Structure Color System Pricing Engine Accessories Audit
OngoingDogfooding

OpenOBA Built by OpenOBA

The OpenOBA platform itself is developed with participation from OpenOBA's own AI Executor. Not a demo—production.

300+ commits
8 releases
2-person team

Challenge

Building an AI Agent OS with complex architecture. A 2-person team couldn't possibly go from zero to V1.6 in 3 months under traditional development.

Approach

The OpenOBA AI Executor participates directly through ERA-Chat—analyzing requirements, writing code, running tests, committing changes. Not "AI-assisted coding"—AI as a team member completing tasks autonomously.

Impact

300+ commits, 8 releases from V1.3 to V1.6 in 3 months. NestJS + Vue 3 + ERDL engine + ReAct reasoning chain—full stack, 1 human + 1 AI Executor. Code pushed to GitHub.

Release Timeline
V1.3.0 → V1.4.0 → V1.5.1 → V1.6.0 …iterating

72 entities × 22 modules backend, 11 views × 18 composables frontend, ERDL engine, ReAct chain, audit—all co-built by OpenOBA's AI Executor. Code pushed to GitHub.

A New Paradigm for the AI Era

OpenOBA + a minimal executable unit. Deploy, start a conversation, tell it what you need. No deep technical background required. No team required. That's your development workflow.

Auth & RBAC Workflows Reports Audit Log Dictionaries Notifications File Mgmt Config

Industry ERDL: eyewear.erdl · apparel.erdl · beauty.erdl — Hot Reload, instant.

See Your Industry in Action

30-minute demo: eyewear ERP + OpenOBA dogfooding dev log. See what an AI Executor can do in your domain.

Request Demo GitHub
Ecosystem

OpenOBA Ecosystem

The AI era offers ISVs and independent developers an unprecedented market opportunity. This is not about AI replacing jobs—it's a paradigm shift that elevates development from coding to higher-value creation. AI never takes your job; it empowers you to turn deep knowledge and experience into sustainable revenue. You just need to shift your perspective.

EDP Market—built on this vision, OpenOBA provides a complete commercial and technical ecosystem: ERDL, Meta-Mirror, and our full suite of proprietary tools. We help ISVs and developers embrace AI by turning processes, practices, knowledge, skills, creativity, and experience into assets—transforming accumulated expertise into sustainable, monetizable products.

Powered by OpenOBA + LLM, you define your own product shape: a full platform, an independent module, a minimal executable unit, or a set of industry practices—all can go live on EDP Market and earn ecosystem rewards.

Join Ecosystem on GitHub

FAQ

Security & Compliance

What if the AI makes a mistake?▼

Every operation passes through the Action Guard protocol layer (valid fields? valid enums?). Everything is recorded in the cognitive audit log. Critical operations require human approval. Like Ctrl+Z—you can always roll back.

Will my data leak to AI providers?▼

OpenOBA provides three layers of protection: ERDL semantic masking, data sanitization on every LLM interaction, and full private deployment support. Your data stays on your servers. For higher security, use your own private LLM deployment.

Can the AI see all my data?▼

OpenOBA has built-in RBAC. Multiple accounts, each with their own scope. Finance sees financial data, operations manage products and orders, warehouse manages inventory—no overlap. Agent permissions are strictly bound to their account's scope.

Cost & Deployment

What's the total cost of ownership?▼

OpenOBA is free and open source (MIT). Your costs: server + model API tokens + optional commercial support. ERDL eliminates semantic ambiguity, reducing token consumption significantly. Compared to traditional outsourcing, OpenOBA reduces TCO by an order of magnitude—not budget cuts, a different production model.

How do I get started?▼

Three steps: Request a demo (30 min), Deploy (Docker, 5 min with our guide), Start describing—tell OpenOBA your first business operation in natural language. Results in 5 minutes. No dedicated dev team needed.

Capabilities

Can it integrate with existing systems?▼

With source code: add 7 standard REST APIs (2-4 person-days). Without: build a lightweight adapter (~200 lines). Or use OpenOBA to rapidly rebuild—add fields, change workflows, connect systems, all in natural language.

How is OpenOBA different from off-the-shelf software?▼

Three differences: OpenOBA gives you a minimal executable unit—auth, permissions, approvals, reports built-in; code and data 100% yours, MIT licensed, no lock-in; modify with natural language—swap an ERDL file, system adapts instantly. One sentence and your AI Executor customizes it.

Which LLMs are supported?▼

Currently default: DeepSeek V4 Pro. Any OpenAI-compatible model (Qwen, GLM, Moonshot, etc.) is plug-and-play. Vision model support coming soon.

About

About OpenOBA

Background

OpenOBA—the best practice of human-AI co-creation, an engineering benchmark for enterprise-grade AI-native execution, and living proof of AI creativity.

Breakthrough

OpenOBA's proprietary ERDL (Enterprise Resource Definition Language)—a protocol-level business semantic description system. Human-readable, machine-readable, readable between machines. Free connection between enterprise digital systems and LLMs, a solid foundation for AI execution.

Core framework open source, full-stack toolchain support, open ecosystem, co-create and share.

Values

Practical Execution

AI shouldn't just chat—it should deliver. OpenOBA puts AI into real business processes, from understanding to database execution, full automation.

Value Creation

Transform human knowledge, experience, and creativity into executable digital assets. AI doesn't replace people—it unleashes higher-value creativity.

Open & Shared

Core framework MIT licensed. Modify freely, use commercially. A minimal executable unit—write your industry ERDL, get a running system instantly.

Human-AI Symbiosis

AI executes, humans decide. Critical operations require human approval. Every step traceable and reversible. AI is an Executor, not a dictator.

Our Beliefs

We Choose Open Source

Not because it's free—because enterprises should own their systems and data. No license expiry, no vendor lock-in. Modify freely, use commercially—forever.

Determinism Over "Smart"

An occasionally clever AI is worthless for enterprise. We need protocol-level validation on every output—not probability, engineering guarantee.

Humans Hold Final Authority

Agents can think and execute, but critical operations require human approval. We're an AI Executor, not an AI dictator.

Ready to Let AI Do the Work?

30-minute demo: eyewear ERP + AI Executor dogfooding. Real cases, no fluff.

Eyewear ERP live on GitHub Full-stack, open source

Request Demo