README.md

metadata

title: Llewellyn Systems Inc
emoji: ⚜️
colorFrom: yellow
colorTo: red
sdk: static
pinned: false

⚜️ Llewellyn Systems Inc

The Operating System for Decision & Enterprise.
Open datasets, open agent primitives, and open infrastructure for the agentic enterprise.

llewellynsystems.com  ·  MCP discovery  ·  agents.json  ·  A2A protocol

---

What we publish here

We use Hugging Face to ship the open layer of our stack — the artifacts an enterprise AI ecosystem can actually be built on.

• Datasets — labeled enterprise operations data, agent-routing corpora, MCP intent traces.
• Models — small, focused classifiers and routers for orchestration (rolling out).
• Spaces — interactive demos of ODE agents, governance probes, and benchmarks (rolling out).

If you train on our data, route through our schema, or extend our agent layer — credit us, ship the work, and tell us about it.

---

⭐ Featured: ODE Enterprise Use Case Dataset

A first-of-its-kind labeled corpus of 15,000 enterprise use cases spanning the full operational surface area of a modern company — procurement, finance, ITSM, supply chain, AI mesh, governance, robotics, and more.

Rows	15,000
Modules	31 (Procurement → AI Policy → Robotics → MDM)
Submodules	215
Verticals	8 (Manufacturing, Healthcare, FinServ, Public Sector, Retail, SaaS, Logistics, Creator)
Channels	5 (Web, Mobile, API, Voice, CLI)
Personas	12
License	CC-BY-4.0 (attribution required)

Built for: training AI orchestrators, intent classification across multi-agent systems, MCP server routing, business process mining, and benchmarking how well LLMs actually understand enterprise operations.

→ huggingface.co/datasets/LlewellynSystems/ode-enterprise-use-cases

from datasets import load_dataset

ds = load_dataset(
    "LlewellynSystems/ode-enterprise-use-cases",
    data_files="use_cases_universal.csv",
)

# Agent routing: map free-text intent → (module, submodule)
for row in ds["train"].select(range(3)):
    print(f"{row['title']}  →  {row['module']} / {row['submodule']}")

---

What we're building

Llewellyn Systems builds ODE — an enterprise operating system designed for an agentic workforce, not a human-only one.

• 19 production MCP servers powering domain agents (finance, procurement, security, ops, AI policy, data lineage).
• 55 AI agent skills orchestrated under a 5-layer governance framework with constitutional AI guardrails.
• Open well-known endpoints — every ODE deployment publishes mcp.json, agents.json, and a2a.json so other agents can discover and negotiate with it.

The artifacts we publish here are the slices of that stack that work better in the open: schemas, datasets, routing models, and reference agents.

---

Citation

If our work shows up in your model, paper, product, or service — cite us.

@misc{llewellyn_systems_hf_2026,
  title  = {Llewellyn Systems on Hugging Face: Open Datasets and Agent Primitives for the Enterprise},
  author = {Llewellyn Systems Inc},
  year   = {2026},
  url    = {https://huggingface.co/LlewellynSystems}
}

---

Get in touch

• Website — llewellynsystems.com
• Engineering / partnerships — llewellyn@llewellynsystems.com
• MCP & A2A integrations — start with the .well-known/ endpoints linked above

Open primitives. Sovereign infrastructure. Built to be built upon.