---
title: Interactive Demo — Cromus AI Workflow Intelligence
description: Try the Cromus 5-step workflow intelligence demo. Score a SOP, compile a SKILL.md, simulate cost across 62 verified models, optimize with Croms™, and export a portable skill pack — all deterministic, no API keys needed.
canonical: https://cromus.ai/demo
source_html: https://cromus.ai/demo
---

# Interactive Demo

> Experience the full Cromus workflow in 5 steps — no account required, no API keys, no external LLM calls.

The demo walks through the complete Workflow Intelligence pipeline on a sample SOP, showing you exactly what Cromus produces before a single token is spent.

---

## The 5 steps

### Step 1 — Score SOP
Paste any standard operating procedure (or use the built-in sample). Cromus returns:
- A readability and AI-readiness score (0–100, e.g. "84 / 100 — AI Ready")
- Findings on structure, role assignments, decision points, and missing fallbacks
- Recommendations for improving AI-readiness before compilation

### Step 2 — Compile SKILL.md
Select an execution target: Generic, LangGraph, CrewAI, n8n, Ollama, or AutoGen. Cromus compiles:
- A portable `SKILL.md` file with token estimates, complexity tiers, and completeness score
- Declared input/output contracts, model requirements, and fallback policy
- Execution profile ready for governance validation

### Step 3 — Simulate Cost
Compare five cost-quality modes side-by-side against the verified model registry (62+ models):
- **Eco** — lowest-cost, smallest models
- **Cost** — cost-optimised with quality floor
- **Balanced** — balanced performance and price
- **Quality** — highest-accuracy commercial models
- **Open Source** — self-hosted, zero-API-cost stack

Each comparison shows per-run cost, 100-runs/day projection, and an EcosystemBadge tier chip on every model.

### Step 4 — Optimize with Croms™
See a Croms™ waste score and an Ops Efficiency Score with concrete optimization actions:
- Downgrade over-specced model tiers
- Parallelise serial steps
- Deduplicate context passed to sub-agents
- Add schema-based retries to flaky prompts

Each recommendation includes an estimated dollar and time saving.

### Step 5 — Export skill pack
Download a ZIP containing:
- `manifest.json` — full workflow spec
- `skills/*.md` — individual skill files
- Wrapper prompts for ChatGPT and Claude
- Ops-savings report (PDF-ready)
- Interoperability JSON for LangGraph, CrewAI, n8n, AutoGen, Ollama, Open WebUI

---

## Deterministic by design

All five steps run with **zero external LLM calls**. Scores, cost projections, and Croms™ values are computed deterministically from the compiled SKILL.md and the verified model registry. The same SOP always produces the same output — no randomness, no model-dependent variance.

---

## Related pages

- [What are Croms™? →](/croms)
- [Pricing →](/pricing)
- [Skill Studio →](/app)
- [SKILL.md Validator →](/validator/skill)