---
title: Open Workflow Ecosystem — Cromus Cross-Stack AI Portability
description: Cromus exports portable SKILL.md workflows to LangGraph, CrewAI, AutoGen, LangChain, LlamaIndex, n8n, Ollama, Continue.dev, and Open WebUI. Build once, deploy anywhere.
canonical: https://cromus.ai/open-workflow-ecosystem
source_html: https://cromus.ai/open-workflow-ecosystem
---

# Open Workflow Ecosystem

> Build once. Deploy to any AI framework. No lock-in.

Cromus compiles SOPs into portable SKILL.md files that export to every major AI execution framework. The Open Workflow Ecosystem is Cromus's commitment to cross-stack portability — your workflows belong to you, not to a vendor.

---

## Export targets

| Framework | Type | Export format |
|-----------|------|--------------|
| **LangGraph** | Orchestration | Python graph definition + node wrappers |
| **CrewAI** | Multi-agent | Crew + Agent + Task config YAML |
| **AutoGen** | Multi-agent | ConversableAgent configuration |
| **LangChain** | Orchestration | Chain + tool configuration |
| **LlamaIndex** | RAG + agents | Pipeline and query engine configuration |
| **n8n** | Workflow automation | Node workflow JSON |
| **Ollama** | Local LLM serving | Modelfile + system prompt |
| **Continue.dev** | IDE agent | `.continue/config.json` configuration |
| **Open WebUI** | Chat interface | Workflow pipeline configuration |
| **Claude Code** | Agent | SKILL.md + system prompt |
| **Replit AI** | Agent | SKILL.md + Replit configuration |

---

## The Open Spec Stack

Portability in Cromus is grounded in three open specifications:

**SKILL.md** — The portable capability file. Any framework that can parse a SKILL.md can consume a Cromus workflow output. The spec is open — any team can adopt it without using Cromus.

**ETHOS.md** — The behavioral governance file. Defines constraints that travel with the workflow across any runtime. An agent that migrates from LangGraph to CrewAI carries its ETHOS.md behavioral policy with it.

**MEMORY.md** — The portable memory spec. Defines what an agent retains across sessions and how memory transfers between agent instances, regardless of the runtime.

All three specs have free public validators at [/validator](/validator).

---

## Why portability matters

AI frameworks evolve rapidly. LangGraph v0.1 to v0.2 introduced breaking changes. CrewAI restructured its agent API twice in 2024. Teams locked into a single framework face migration costs every time the framework changes.

Cromus workflows are framework-agnostic. The compiled SKILL.md is the source of truth — framework-specific exports are generated from it on demand. When a framework changes, you re-export from the same SKILL.md rather than re-architecting the workflow.

---

## Open-source model support

The Open Source cost-quality mode in Cromus simulates workflows against self-hosted model stacks:
- **Ollama** — local model serving (LLaMA, Mistral, Qwen, DeepSeek, Phi)
- **Open WebUI** — self-hosted chat + pipeline interface
- **Continue.dev** — IDE agent with local model backend
- **LM Studio, Jan, GPT4All** — local inference clients (SKILL.md compatible)

Self-hosted stacks eliminate per-token API costs entirely. Cromus quantifies the quality trade-off using verified open-source model benchmarks.

---

## Related pages

- [AI Workflow Templates →](/ai-workflow-templates)
- [AI Skill Packs →](/ai-skill-packs)
- [Low-Cost AI Workflows →](/low-cost-ai-workflows)
- [SKILL.md Validator →](/validator/skill)