Open kimono · Built in public

This is what
AI-native looks like
in production.

VibeTokens runs on Claude. Not as a productivity tool — as infrastructure. Every system documented here is live, running, and costing us fractions of a cent per operation. We're showing all of it.

$0.07

Per full ICP outreach run

28

Agent roles defined in the org chart

6am

CEO brief delivered every morning

100%

Content engine zero-touch

Daily at 6am

CEO Intelligence Brief

Running
Claude APIGmail APIGoogle CalendarSQLite

What it does

Pulls email, calendar, open pipeline, and relevant industry news. Claude synthesizes it into a five-minute morning brief and delivers it to inbox. No apps visited. No context-switching. The day starts with everything already assembled.

Why it exists

Gloria Mark at UC Irvine put the recovery time from a single interruption at 23 minutes. The brief eliminates eight context switches before 9am. That's roughly two hours of cognitive overhead, daily.

Nightly at 2am

Outreach Pipeline

Running
Claude APISQLitePlaywrightGmail SMTPAnthropic SDK

What it does

Discovers ICP-fit prospects from public sources, scores them against criteria, drafts personalized cold emails, and schedules delivery. Runs autonomously. One command starts the cycle. No human touches the outreach until a reply comes in.

Why it exists

At $0.07 per full ICP run, the economics are different from traditional outreach. The constraint isn't cost — it's quality and timing. The system handles both.

Daily at 7am

Content Engine

Running
Claude APINext.js MDXGitHub APIVercel

What it does

Selects a topic from a rotating research queue, drafts a full blog post as MDX, writes the frontmatter, commits to the vibetokens GitHub repo, Vercel deploys automatically, and stages a LinkedIn adaptation in a pending file. One npm command, zero humans.

Why it exists

Content compounds. A post from six months ago still drives traffic. Building the engine once and letting it run produces that compounding without the bottleneck of someone having to sit down and write.

On demand

Revenue Tracker

Running
SQLiteTypeScript CLIbetter-sqlite3

What it does

CLI-native MRR tracking. Log a deal, view the pipeline, get the current MRR number. No spreadsheet. No Notion database. No dashboard that takes four clicks to load. The number is one command away.

Why it exists

Revenue visibility shouldn't require a dashboard. When the number is in a database and the CLI is in your terminal, you check it. When it's in a spreadsheet someone updates manually, you estimate.

Post-publish

LinkedIn Engagement Layer

Running
PlaywrightChrome CDPTypeScript

What it does

After the content engine stages a LinkedIn post, a Playwright script connects to the existing Chrome session via CDP and posts it directly. No third-party scheduler. No API token. Just Claude's draft, posted from the actual account.

Why it exists

LinkedIn's algorithm rewards native posts over scheduled-tool posts. Using CDP means no footprint difference from manual posting.

Why we publish this

The consulting pitch for AI is full of theory. We wanted ours to be full of receipts. Every system on this page is running. The costs are real. The timelines are real. The code is on GitHub.

We're not showing you what's possible. We're showing you what we've already done — and what we can build for you if your operation is ready for it.

The architecture compounds. The systems get better with use. The context gets richer over time. The gap between businesses that build this way and businesses that don't gets wider every month.

Build yours

Tell us what your
operation should look like.

We scope every engagement to the specific operation. Tell us how yours runs and what you'd change — we'll come back with what we'd build first.

MURPH · VIBETOKENS INTAKE