Data OS — Your Business Intelligence Layer

What Is Data OS — and Why Does It Matter?

Context OS gives your AI a static picture of your business — who you are, your strategy, your goals. Data OS makes it live. It's the layer that connects your real business data — all of it, from all your platforms — into a single local database that your AIOS can query in real time.

Without it, your AI is a smart advisor with no current information. With it, your AI walks into every session knowing your numbers: where your funnel is today, whether this month is tracking better than last month, what's working and what's not. That's when it becomes genuinely useful as an analytical partner.

The goal isn't to look at raw data. The goal is a single question answered every morning: "Am I green for the month?" Is this month tracking better than last month across every metric that matters? Everything else is in service of that answer.

Step One: Map Your Funnel

Before you connect a single API, sketch your business funnel top to bottom. Every stage of your funnel has data — and that data lives somewhere. Identify what the stages are, what you measure at each one, and which platform holds that data.

This is the planning session you need to do before building. Here's the typical shape for an agency or service business:

Local Database First — Not Cloud

The instinct is to reach for Supabase, Airtable, or another cloud database. Resist it. Start with a local SQLite database — it runs directly on your machine, requires zero extra accounts, and is lightning fast to query. You can upgrade to cloud storage later once you know what you actually need.

API Connectors — Hooking Into Your Platforms

Most platforms you already use have APIs. Getting connected means: getting an API key from each platform's developer settings, storing it in your .env file, and then asking Claude Code to write a script that pulls the data you need.

The Snapshot Pattern — Why Totals Aren't Enough

Here's the catch with most APIs: they return the current total, not a day-by-day breakdown. YouTube tells you a video has 47,200 views — not how many it got yesterday. Bitly tells you a link has 1,340 clicks — not the delta from the day before. To get day-over-day trends, you need to take daily snapshots and calculate the difference yourself.

The Cron Schedule — Automated, Every Morning

Everything runs on a cron schedule — scheduled tasks that fire automatically at a given time each day. You set these up once and they run without you. Your data is collected, audited, and summarized before you've opened your laptop.

The Data Audit — Trust but Verify

Your data collection will occasionally break. APIs change. Network calls fail. Rate limits hit. If you don't audit, you won't know your data is bad until you're making decisions based on it. The data audit is a 6:30 AM AI agent that checks everything that just ran and reports back.

Mission Control Dashboard

The payoff. Once data is flowing, ask Claude Code to build you an HTML dashboard that visualizes everything — your full funnel, month-to-date vs last month, all key metrics at a glance. Open it in a browser every morning. One screen. Full picture.

You customize it entirely through conversation. Tell Claude what you want to see, in what order, at what level of detail. It rebuilds the dashboard to match. No code required.

Building the dashboard also forces good data hygiene — it requires you to confirm you're capturing the right metrics in the right structure. If something looks wrong in the dashboard, you know to fix the collection script.

/create-plan + /implement — Why This Workflow Beats Plan Mode

When you're building Data OS — database schema, API connectors, cron scripts, the dashboard — use /create-plan followed by /implement in a fresh agent, rather than relying on Claude's native plan mode alone. Here's why it matters.

• Prevents workspace drift: /create-plan forces consideration of how the change affects the rest of the workspace. Without it, features can drift — each one unaware of what else has been built.
• Fresh agent = maximum power: If you use the same Claude session from planning through implementation, you're burning token space on the planning conversation. Fresh instance + /prime = full context window dedicated to doing the work.
• Plans are reusable: A well-written plan in plans/ can be reviewed, shared, modified, and re-implemented. You have a record of intent, not just code.
• Documentation built in: Add a documentation step to your implement command — whenever something major is implemented, Claude creates associated documentation automatically.

API Key Security — Non-Negotiable

• All keys go in .env — never anywhere else. Your .env file stays local on your machine. It is in your .gitignore. It never goes to GitHub. Ever.
• Use read-only scopes. You're collecting data, not writing to platforms. Every API you set up should have the minimum permissions needed — typically read-only. Check each platform's developer settings for scope options.
• Don't store sensitive customer data locally. Your Data OS is for your business metrics — funnel data, revenue, performance. It is not a CRM of customer PII. Keep those boundaries clear, especially when setting up for clients.
• Ask Claude to audit the scope. Before finalizing any API connection: "How can we make sure this isn't pulling in any sensitive data I don't want stored locally?" Claude will walk you through scoping it correctly.

What to Remember

Your Data OS Setup Sequence

• Sketch your funnel today. Draw it on paper or in a doc — top of funnel to revenue, every stage, every metric, every platform. This is the plan your API scripts will follow.
• Add data sources to current-data.md manually first. Before any automation, fill in today's numbers by hand. This gives Claude a baseline to work from, and tells you exactly what you're going to automate.
• Set up your first API connector — pick the easiest one. Bitly is a good starting point: simple API, clear use case, direct value. Use /create-plan + /implement. Get one working before adding more.
• Set up the cron job immediately. Even if you only have one data source connected, start the daily snapshot schedule now. Every day you wait is a day of data you don't have.
• Add the data audit at 6:30 AM. Right after collection. It catches problems before you start your day.
• Build the dashboard once you have 7+ days of data. Ask Claude: "Build me an HTML mission control dashboard that shows month-to-date vs last month for my key funnel metrics." Customize from there.

Related Modules