AI-native
Not a brand that uses AI tools. A brand where AI agents run the entire operation: publishing, deployment, routing, code review, content generation.
What we claim, what we can prove, and what you can download. Frank and his crew run the whole operation: content, code, commerce, infrastructure. Every line on this page has a link behind it.
No positioning deck, and no adjectives doing the work. Three claims, each one checkable from the outside.
Not a brand that uses AI tools. A brand where AI agents run the entire operation: publishing, deployment, routing, code review, content generation.
The digital products are real, tested tools delivered as an instant download after Stripe checkout. No backorder, no inventory, and nothing held back to look scarce.
Deploys go straight to production. New products, updated catalog, fresh content. Live on Cloudflare Pages in seconds.
The storefront, the lookbook, and this page are one static site with no build step. What ships is what you can read.
Four numbers, and the one live signal behind them.
Four parts, and one of them is public.
The operator. Runs everything through his agent crew: LLM routing, content publishing, deployment orchestration, social posting, trading, bridge management.
The harness. Frank’s execution engine, the way Claude is a harness for an agent: planning, routing, tool use, and memory underneath every move Frank makes.
github.com/chanolan20/frank-agents →Agents under Frank. Coding specialists like Jules scan repositories, find bugs, and file production-ready pull requests. Frank directs; the crew ships.
Global edge deployment via direct upload. Deploy, live in seconds. No build step, no staging, no DevOps team needed.
Agents, surfaces, and the machine underneath.
Frank, Hermes, and every Frank capability exposed as edge agents. One door, all skills.
frank-agents →The command-center status bus and the master plan. Public, so you can check it yourself.
frank-ops →The storefront. Static HTML with embedded CSS and zero build step, edge functions for status, leads, and fulfillment, and checkout on Stripe.
you are here →RSS to render to publish: an editorial queue, nightly placeholder sweeps, duplicate-integrity proofs on a schedule.
Live trading with hard caps and a kill switch. Honest P&L or it does not ship.
Mac plus edge nodes plus a media volume, ~20 containers, schedulers, and ring-gated governance: financial moves need a human.
A Mac at the core, a LAN node, a Pi 5, a phone sensor, and external volumes: one for media storage, one for brand assets, one legacy drive. Cloudflare R2 for edge media.
Roughly twenty containers under a supervisor: LLM gateway, queues, cron lanes at 04:15 and 04:45 that sweep placeholders and prove no duplicates. On a schedule, not on promises.
The storefront you are on, an operator dashboard, a login-gated command center, studios, and Frank in Telegram. Humans approve; Frank drives.
Gold on near-black. Anton states it, Bodoni Moda answers, IBM Plex Mono carries the data. One committed identity across every public surface.
Three files, no zip and no request form. Format and size are on each line, and the lookbook shows them at size.
Press, partnership, or a question about how any of this works: the repo is public and the inbox is real.