About Cromus
Most organizations running AI workflows don't have an AI department. They have engineers who build, founders who decide, and spreadsheets that try to track what it all costs — with no dedicated function watching governance, drift, or spend.
Cromus is the AI department you don't have to hire.
Three roles make up the department: the Specialist, the Manager, and the CWO. Builder ($49/month) starts you with the Specialist — your always-on workflow reviewer, checking every SKILL.md change against 24 governance rules before it ships. Ops adds the Manager, who watches cost drift and Croms accumulation across your entire stack. Agency adds the CWO, who signs the governance report leadership and clients need to see.
One function. No headcount required. Free to try — no credit card.
The Three Roles
The Specialist
Reviews every SKILL.md change in CI. Catches model mismatches, missing governance fields, and structural gaps before a single token is spent.
The Manager
Watches the stack continuously. Tracks cost drift, latency overhead, and Croms accumulation across every workflow in the workspace.
The CWO
Signs the governance report. Surfaces the ETHOS.md decisions, risk flags, and compliance posture that leadership and clients need to see.
What Cromus Does
Cromus is the pre-execution intelligence layer for AI workflows — the platform that scores, governs, and makes workflows portable before tokens are spent. It is the AI department that sits between workflow design and workflow execution: the function that answers questions about cost, governance, and accountability before they become incidents.
Every SKILL.md file is a governed, portable workflow specification. Every ETHOS.md file is a behavioral governance contract. Every workspace comes with a Croms score — a real-time measure of preventable waste across cost, latency, failure risk, and structural gaps. Total Cost of Workflow Ownership (TCWO) translates those scores into dollar terms over time.
The platform is AI-agnostic and ecosystem-portable. Workflows are governed without being locked into a specific model, provider, or runtime — and can be compiled to LangGraph, CrewAI, n8n, local stacks, and more.
Who Cromus Is For
- Teams running AI workflows who need cost accountability without hiring a dedicated AI ops function
- Developers and engineers who want SKILL.md governance baked into CI from day one
- Founders and product leads who need to show AI ROI without spreadsheet archaeology
- Agencies and consultancies managing AI workflows across multiple client stacks
- Researchers and analysts who want a portable, auditable record of how their workflows behave
Open Standards at the Core
The Cromus platform is built on open specifications: SKILL.md for portable workflow definitions, ETHOS.md for behavioral governance, MEMORY.md for portable agent memory, and GRAPH.md for workflow topology. Each specification has a free public validator. The scoring rules and evaluation logic are published openly so teams can adopt the standard independently.
Cromus maintains a verified model registry covering 96 active generative models from 13 text-generation providers, plus TypeSafe's input-priced Jev 1.13 decision model (14 providers total) — OpenAI, Anthropic, Google, xAI, DeepSeek, Mistral, Meta, Cohere, Amazon, Microsoft, Qwen, Z.ai, Kimi, and TypeSafe — so every cost simulation reflects current, real pricing.
The goal is a world where AI governance is not an afterthought bolted onto a workflow after it's already burning budget — but a first-class function, staffed from day one, at a price any team can afford.
Our mission
The AI stack has a blind spot. We're naming it — and filling it.
Execution tools run workflows. Agent frameworks orchestrate them. Observability reports what they cost after the fact. But the moment before — where architecture, model choice, and budget get decided — has no layer and no name. Teams make their most expensive AI decisions in the one place the stack can't see.
Cromus closes that gap. The pre-execution layer that scores, simulates, and governs AI workflows before a single token is spent — because cost transparency shouldn't be a post-mortem.
See how pre-execution governance works — and what the Croms data says about where AI budgets actually leak.