How We Built Awen OS: The AI Operating System Behind Our Studio
Two people run the output of a much larger team. The reason is an AI operating system we built ourselves and never sold. Here is exactly how it works.
Most studios run on a dozen disconnected tools: a CRM here, a newsletter app there, a project board, a social scheduler, a docs folder, and a pile of spreadsheets holding it together. We replaced all of it with one system. Awen OS is the AI operating system we built to run every brand we operate, from a single place, with an AI assistant woven through every workflow.
One system. Every brand. AI in every workflow.
Why we built it
Awen is not one company. It is a federated studio: a record label, an artist roster, a Web3 consultancy, a women's blockchain community, a youth soccer nonprofit, and more. Each has its own audience, brand, and pipeline. Off-the-shelf tools assume one company doing one thing, so nothing fit a multi-brand creative studio that needed shared infrastructure without blurring the brands together.
The problem with off-the-shelf
Run eight brands on standard software and you get eight of everything: eight logins, eight subscriptions, eight exports that never quite line up. Worse, the tools fight your structure. A CRM wants one company. A newsletter platform wants one list. A scheduler wants one set of accounts.
We needed the opposite: one backend that knows every brand is distinct, keeps their data and identity separate, and still lets a two-person team move across all of them in a single day.
One hub. Every brand.
Hub-and-satellite: each brand is its own node with its own identity, all sharing one core.
How it is built
Awen OS runs on WordPress core with a custom REST and JWT layer, so every screen is an API call and the whole thing installs as a progressive web app at app.awen.online: native-feeling, offline-capable, with push notifications.
The model is hub-and-satellite. Each brand is a satellite that owns its identity (logo, colors, voice, socials) and pushes it to a central hub registry. The hub mirrors that identity out to every surface, the website footer, the email signature, the social cards, so a brand is defined once and stays consistent everywhere.
Six modules, one system
- 01
CRM
Contacts, deals, and pipeline across every brand, with activity logging and per-brand scoping.
- 02
Tasks
A shared board with per-person daily and sprint views, rich-text descriptions, and notifications.
- 03
Newsletter & Social
Campaigns with open and click tracking, plus one post fanned out to every channel in the right brand voice.
- 04
Docs & Pulse
Branded quotes, invoices, proposals, and surveys, plus analytics that turn raw data into one read.
Agentic by design
AI here is not a feature bolted onto a settings page. It is the connective tissue, and increasingly it is agentic: the same assistant drafts a newsletter, triages a new lead, summarizes a meeting into tasks, and dispatches the daily status of a project, because it can see the whole system at once. That is context engineering in practice, the model works because it has the full picture, not a clever one-off prompt.
What the assistant does
Drafts newsletters and social posts in each brand's distinct voice. Triages and enriches new CRM contacts. Turns meeting notes into tasks, documents, and follow-ups. Generates a daily pulse on every active project so nothing goes quiet.
Why it matters
The point is not novelty, it is leverage. Every workflow the agent touches gets faster, so the humans spend their hours on the work that actually needs judgment, taste, and relationships.
Sovereign by default
Awen OS is not a stack of rented SaaS subscriptions. We own the system, the data, and the way the AI is wired into it. In a year when sovereign AI and data ownership jumped from buzzword to board-level priority, our studio already runs on infrastructure we control end to end.
Proof across the studio
The same system runs very different work. For the record label, it tracks years of sales and turns Bandcamp purchases into ready-to-post stories. For the blockchain community, it manages member segments and sends legislative invitations to the right people. For the music-licensing platform, it helps answer incoming sync briefs from a song catalog. For the soccer nonprofit, it drafts and sends the monthly invoice on schedule.
One backend. One agentic assistant. Many businesses.
What it means for you
We have lived every edge case of putting AI into real, daily work. That is the experience we bring when we build for other teams: custom AI tools, internal platforms, client portals, and Claude-powered agents that operate your systems directly over MCP, designed around how you actually work.
Where it goes next
Some of what used to be on this roadmap now ships. Awen OS speaks the Model Context Protocol, the open standard from Anthropic now stewarded by the Linux Foundation: seventy-seven of the operating system’s actions are exposed as agent tools, so an assistant can operate it directly instead of only drafting inside it. What keeps that safe is a three-tier permission model — read, draft, and commit — where anything that reaches the outside world stops for explicit human confirmation first.
What is still ahead: multi-agent orchestration, where specialized agents handle research, outreach, and production in parallel under one coordinator, and assistants that do not just suggest the next step but take it. The system that runs our studio is becoming the template for the agentic systems we build for clients.
We build agentic AI systems for other teams.