Arcanian OS

An open-source go-to-market diagnostic system.
Find which layer of your marketing is actually broken — not just the channels.

7 layers. 37 skills. 14 AI agents. Built by a 3-person team diagnosing go-to-market problems for 17 clients across multiple markets.

Most companies optimize channels. The constraint is usually deeper — the offer, the customer definition, the positioning. This system finds it.

We open-sourced the methodology. Not the client data. The thinking that makes go-to-market decisions repeatable.

Powered by Claude Code and Databox MCP.

View on GitHub Getting Started Guide

The core: a 7-layer go-to-market diagnostic

  1. L7MarketAre we even in the right game?
  2. L6CustomerDo we actually know who buys?
  3. L5ChannelsAre we visible where it matters?
  4. L4OfferDoes what we sell match what they need?
  5. L3PositionWhy us and not them?
  6. L2IdentityWhat does our brand actually say?
  7. L1CoreCan the team actually execute this?
  8. L0SourceWhat do we believe about marketing?

What's in the box

  • 37 Skills

    Diagnostic routines: 7-layer analysis, measurement audit, GTM analysis, health checks, copy analysis, content quality. Each one a structured prompt that loads context, applies a framework, and produces actionable output.

  • 14 AI Agents

    Specialized workers: audit checker, channel analyst, PII scanner, report reviewer, data rules enforcer, knowledge extractor, and more. Each has a defined role, frontmatter, and guardrails.

  • 4 Council Configs

    Multi-agent deliberation. When the system is uncertain, it spawns a panel of agents to debate the same data from different angles — generates competing hypotheses, scores evidence, produces consensus.

  • 27 SOPs

    Standard operating procedures for marketing ops, measurement audits, client onboarding, campaign management, agency coordination, inbox management. The playbook that makes execution deterministic.

  • Databox MCP

    Pull metrics via natural language. Domain-filtered queries across Google Ads, Meta, GA4, Shopify, ActiveCampaign. The system loads the right data source, applies reliability ratings, and runs analysis frameworks automatically.

  • Guardrails

    Data reliability framework. Evidence classification. Domain isolation. Discovery-not-pronouncement — the system helps you think, not tells you what's true. PII scanning on every output.

  • Scaffold

    Everything you need to start a new project: CLAUDE.md, TASKS.md, CLIENT_CONFIG, audit templates, deliverable routing, contact registry, project glossary. One command sets up the entire structure.

  • GTM Gap Analysis

    Six core go-to-market decisions: market, early customer, product, pricing, positioning, growth. The /analyze-gtm skill runs the full framework and shows which decisions are missing or broken. One command. The strategic layer most dashboards never touch.

How it works

Your CLAUDE.md is the kernel. It loads skills, routes to SOPs, enforces guardrails, and connects to Databox for live data. The file IS the operating system.

The 7-layer diagnostic drives everything. Every engagement starts with /7layer. It maps where the marketing system is broken — not just the channels, but the beliefs, capabilities, positioning, and offer underneath. The rest of the system exists to act on what it finds: /repair-roadmap builds the fix plan, /measurement-audit checks the tracking, /health-check monitors ongoing.

Agents do the watching. PII scanner checks every output before it reaches a client. Audit checker looks for measurement gaps. Data rules enforcer accounts for sync delays before flagging anomalies.

The council resolves uncertainty. When something looks off — a 74% drop in conversion value, a sudden traffic spike, a metric that doesn't add up — the system doesn't guess. It spawns agents to generate competing hypotheses, score each against the evidence, and produce a ranked assessment.

Databox is the data layer. 130+ integrations aggregated in one place. The MCP server lets your agents pull metrics in natural language. No manual dashboard building. No CSV exports. Ask the question, get the analysis.

Who this is for

You run marketing across multiple clients. And you're tired of rebuilding the same analysis framework for each one.

You have Databox connected. And you want AI-powered analysis on top of your data, not just dashboards.

You manage multi-domain operations. And you need domain isolation, not account-level totals that mix everything together.

You believe marketing ops should be systematic. Not tribal knowledge in someone's head. Not a Notion doc nobody reads. A system that loads the right process at the right time.

What's next

This system is in active development. Here's where we are and where it's going.

Shipped

  • Confidence engine — every finding scored by source reliability, evidence class, and assumption status
  • Per-platform channel analysts — dedicated agents for Google Ads, Meta, and Shopify
  • Baseline detection — the system learns what "normal" looks like per client and flags real changes, not sync delays
  • Business unit isolation — multi-brand clients get clean numbers per unit, not blended totals
  • Attribution window awareness — cross-platform comparisons flag measurement mismatches automatically
  • Recommendation tracking — the system checks whether past advice actually worked before recommending it again
  • Currency normalization — cross-market analysis converts before summing

In progress

  • Walkthrough videos — Sales Pulse, measurement audit, council deliberation, and morning brief on live demo data
  • Training materials — getting from clone to first /health-check in 30 minutes
  • GTM strategy content series — showing how /analyze-gtm and /7layer diagnose go-to-market gaps
  • Demo Databox environment — synthetic data with planted anomalies so you can test the system without connecting your own accounts

Coming next

  • Outcome effectiveness dashboard — Databox visualization of which recommendations moved the numbers
  • Setup session offering — 2-hour guided configuration for agencies who want help getting started
  • /diagnostic page on arcanian.com — interactive 7-layer framework with self-assessment
  • Community contributions — accepting PRs for new skills, agents, and platform-specific SOPs

Ideas we're exploring

  • Slack bridge — morning brief and anomaly alerts delivered to a channel
  • More platform agents — TikTok Ads, Pinterest, Amazon, LinkedIn Ads
  • Client-facing reports — auto-generated deliverables from council outputs
  • Multi-language support — English and Hungarian SOPs and diagnostic skills

Working with

  • Databox
  • ActiveCampaign
  • Semrush
  • Amazee.io

Need help getting started?

The repo is public on GitHub. If you want help configuring it for your setup — or want to talk through how it fits your operation — tell us about your situation.

Start a conversation