---
title: Cromus for AI Engineers — Workflow Governance and Cost Simulation
description: AI engineers use Cromus to compile SOPs into SKILL.md, simulate cost across 62+ models, score Croms™ waste, and export to LangGraph, CrewAI, n8n, AutoGen, and Ollama before any execution.
canonical: https://cromus.ai/for/ai-engineers
source_html: https://cromus.ai/for/ai-engineers
---

# Cromus for AI Engineers

> The pre-execution intelligence layer for AI engineers — cost simulation, waste scoring, and portable compilation before any execution.

AI engineers build the systems that run AI workflows. Cromus is the tool for the design and governance phase — before any code is committed to a specific model or framework.

---

## What AI engineers use Cromus for

### Model selection before coding
Compare 62+ models across 11 providers — with verified, weekly-updated pricing — before writing any integration code. Avoid committing to a model that's 10x more expensive than necessary for the task complexity.

### Framework-portable compilation
Compile a SOP into a SKILL.md once. Export to LangGraph, CrewAI, AutoGen, LangChain, LlamaIndex, n8n, Ollama, Continue.dev, or Open WebUI on demand. Change frameworks without re-architecting.

### Croms™ waste scoring
Get a deterministic waste score for any workflow. Identify tier mismatches, parallelizable steps, context bloat, and flaky retry policies before they become production issues.

### ETHOS.md and MEMORY.md governance
Declare behavioral policies (escalation rules, refusal conditions, transparency requirements) and memory contracts (retention scope, compression strategy) alongside the capability spec. Governance travels with the workflow across any runtime.

### MCP server integration
The Cromus MCP server at `https://mcp.cromus.ai/mcp` exposes six deterministic tools (`score_skill`, `simulate_cost`, `validate_skill`, `validate_ethos`, `validate_memory`, `reality_check_summary`) callable from Claude Cowork, OpenAI Codex CLI, or any MCP-compatible client. Integrate workflow intelligence directly into your engineering workflow.

---

## Technical details

- All scoring and simulation is **deterministic** — no LLM calls, pure functions of the SKILL.md and model registry
- Model registry covers **62+ models**, **11 providers**, with **weekly pricing and deprecation checks**
- SKILL.md, ETHOS.md, and MEMORY.md are **open specs** with free public validators
- MCP server supports **OAuth 2.1 + PKCE** for team deployments
- MCP server operates under a **no-retention contract** — inputs are not stored or used for training

---

## Export formats

| Framework | Export format |
|-----------|--------------|
| LangGraph | Python graph definition + node wrappers |
| CrewAI | Crew + Agent + Task config YAML |
| AutoGen | ConversableAgent configuration |
| n8n | Node workflow JSON |
| Ollama | Modelfile + system prompt |
| Continue.dev | `.continue/config.json` |
| Open WebUI | Pipeline configuration |

---

## Related pages

- [Cowork Connector →](/cowork-connector)
- [Codex Plugin →](/codex-plugin)
- [SKILL.md Validator →](/validator/skill)
- [Open Workflow Ecosystem →](/open-workflow-ecosystem)
- [Pricing →](/pricing)