Everything you need to know before you start.

Founder-led service businesses with 10 to 30 people, mostly in the UAE and the wider GCC. AED one to ten million revenue is the centre of gravity. You already pay for AI, already have a team, and are tired of being the answer to every question. If that sounds like your week, ARCAS was built for you.

We give founders a second brain their team can think with when the founder is not in the room. The free piece is Founder OS, a folder that gives your AI your business context. The paid piece is setting that brain up on your team's side so your judgment, your decisions, and your operating context become something the team can query instead of asking you.

A consultant arrives with a deck and leaves you with a deck. ARCAS arrives with a working brain and leaves you with a working brain. We do not bill by the hour, we do not run open-ended retainers, and we do not hand over slides. The deliverable is a system your team uses on Monday, not a strategy your team forgets by Wednesday.

Both, plus open source. Founder OS lives on GitHub. Clone it, run it on your machine, never speak to us, that is a valid path. The paid engagement is for founders who want the brain set up on the team side, populated with real operating context, and kept tuned as the business changes. Self-serve is enough when you have the time and the discipline. The engagement is for founders who would rather buy back the install week.

Because adding AI to a business with unclear roles, undocumented decisions, and a founder bottleneck does not give you leverage. It gives you faster chaos. People and systems are what AI runs on. We sequence the work the same way every time.

A folder of files that gives your AI your business context. Plain English text that your AI reads before it answers you. Your decisions, your team, your week, your network, your operating rules. The format is markdown, you do not need to know what that means. Lives on your machine. Your data, your folder, your control.

No catch on the ARCAS side. Clone from GitHub and there is no gate at all. The quick download asks for your email so we can send setup instructions, that is the only ask. No account. No analytics watching what you put in the folder. No way for us to see your data even if we wanted to. Delete the folder and the OS is gone.

Founders who get value out of Founder OS often want help rolling it out across their team. That is where ARCAS gets paid. The repo is how the right partners find us. If you install it, get value, and never speak to us again, that is a fine outcome too. We would rather build trust than charge for a download.

The folder is on your computer. We never see it. It is not synced to our servers. When your AI reads from it, that content goes to Anthropic or OpenAI, the same as anything else you type into a chat. The difference is you own the folder, not us. Back it up to your own private cloud or Git. Delete it in one step. Nothing is locked away from you.

Day one you stop explaining yourself for the same task twice. Day three your AI plans your week without being briefed. Day seven your weekly review writes itself and your decisions stop disappearing. Most founders feel the shift inside the first week.

Around USD 20 a month to Anthropic or OpenAI for the AI side. Nothing comes to ARCAS. See the flow on /founder-os for the full picture.

Claude Code and Codex can read files on your machine directly. Your AI just opens the relevant section of your folder when it needs it. The regular ChatGPT and Claude.ai chat apps cannot read your files, so you have to paste in the relevant section at the start of every conversation. Same idea, more friction, more chance of forgetting. The difference between fifteen seconds of context-setting per chat and zero shows up by the end of week one.

Works, but slower. The free plans hit message limits faster, and they do not give you Claude Code or Codex, so you stay on the copy-paste workflow. Most founders who try Founder OS on free plans end up upgrading to one of the paid plans within two weeks because the friction is what kills the habit. Try it on free first if you want, then decide.

No. The folder is yours. Built on Claude today, decide tomorrow that Codex is sharper, every file you wrote transfers, no migration, no export, no permission to ask. The lock-in problem is the part of the AI market we built around, not the one we participate in.

No. Zero affiliate, zero kickback, zero per-seat resale. You pay Anthropic or OpenAI directly for your AI subscription. We pay them too, on the same plans, like everyone else.

Three places, in order. First, anything where the brain does not hold the context yet. Early in the install your AI will guess and sound confident doing it. Run /founder-os:status, the gaps surface, you fill them, the hallucinations stop. Second, anything judgment-heavy where the same question can come back two different ways depending on phrasing. Pricing a deal. Reading a senior hire. Calling a client risk. The AI drafts, you decide. Third, anything that needs reading a person in the room. Body language. The team member who is unhappy and not saying it. The brain captures patterns you already see. It does not see what you cannot see.

Modern AI is non-deterministic by design. Ask it the same judgment question twice and you can get two valid framings, especially when context is sparse or the question is open-ended. The fix is on the brain, not on the AI. Tighter context, tighter prompts (which is what the slash commands are), tighter decision rules. The closer the brain matches how you actually decide, the less the variance matters. Treating the AI output as a draft you sharpen, not an answer you accept, is the discipline that keeps you in control.

No. Not in the first month, probably not ever for the calls that move money or people. The OS drafts. You decide. It writes the proposal, you read it. It plans the week, you push back. It surfaces the call a client meeting needs, you make it. The day you let it run autonomously is the day you stop being the founder. That is not the trade we are offering.

No. The diagnosis engine is rule-based and deterministic. Same answers in, same results out, every time. We generate AI-ready prompts for your use, you copy them into the AI you prefer, we never send your business data to an AI provider.

A free stress test. Five to eight minutes, seven areas, scored output. The report maps your leaks across the Five Levels (shown above). One route into ARCAS. Others skip it and start with Founder OS. Either entry point works.

No. Free. No login. No paywall. When you download or email the report, we ask for your email once so we can send it. No account is created.

Five to eight minutes for the full scan. The engine recommends the focus areas most likely to matter for your business. You can switch areas at any point. Progress saves automatically.

A branded PDF report, AI-ready prompts you can copy into any LLM, a one-week focus plan, and a scored breakdown of where the business is leaking. After that, three paths. Self-serve with the playbook and free tools. Book a thirty-minute call to pressure-test the findings. Or move into a paid engagement if the fix needs hands-on help. You decide.

Book the call. Bring the report. We pressure-test the findings together. The engine is deterministic, the same answers always produce the same results, but context matters and the call is where context lands.

Thirty minutes. We drop your public context (LinkedIn, company website, anything you share) into the second brain live. You watch your own voice and your own patterns come back at you, and the shape of a team-side question your brain could already answer. Not a sales pitch. Not a needs assessment. The brain working on your business in front of you so you can decide if the engagement is worth it.

Yes. Always. No card, no upsell flow. If we are not a fit, you walk away with what you saw and a clearer picture of where the business actually leaks.

The person who built the engine. No account managers, no handoffs.

No. Some founders prefer to see the brain working before they answer questions. Others want the report first. Both routes lead to the same place.

Your team running on the same brain you are. We set the brain up on the tools your team already uses (Notion, Slack, WhatsApp, whichever stack), populate it with your real operating context, train one internal owner, and run the first weekly rituals with the team in the room. Where adoption drops, we go back in. The brain is a living thing, not a deliverable that gets handed over and forgotten.

Scoped on the call against a hard commercial target we agree at the start. You see the full number before you commit, not after. We do not bill by the hour, there is no open-ended retainer, and pricing does not appear on the public site because the right answer depends on what you actually need.

Decision load drops first. The team queries the brain before they query you, so the volume of questions hitting your phone falls. Delivery follows a system instead of one person's memory, so quality across clients evens out. Onboarding compresses, because the brain holds the context a new hire would otherwise spend months absorbing. By the end of the engagement, the recurring work runs without the founder in the room.

Founder-side work is fully remote. The team-side engagement embeds on-site for the parts that need it, usually the first two weeks and again around major milestones. We have to see how the team actually operates, not how the founder describes it. Outside the UAE and GCC, we can run remote-first, with results that depend on how willing the team is to be observed on video.

Founder OS engagement: roughly fifteen to thirty days for the focused install. Team-side engagement: roughly sixty to ninety days for the rollout. The retainer tier is monthly after that, optional, designed for businesses that want the brain to keep up as the company evolves.

You can. Engagements are scoped in phases and each phase ends with a written review against the commercial target we agreed. If the numbers are not tracking, we tell you before you tell us. You keep everything built up to that point. No retainer to exit.

Adoption is the engagement, not a side effect. We train one internal owner, run the first weekly rituals with the team in the room, and review usage weekly. Where adoption drops, we go back in. If the team will not query the brain, the brain is useless. The retainer tier exists for exactly this reason.

The free repo gives you the chassis and one founder running a sharper week. The paid engagement gives you the team running on the same brain, the install done correctly the first time, and someone keeping the brain alive as the business changes. Information is free. Implementation in your actual team is what produces the change.

No. Founder OS lives on your machine. The diagnosis stays in your browser by default. We do not sell data, do not use diagnosis data for sales calls, and do not log what you put in the brain.

Your folder lives locally. When your AI reads from it, the content goes to Anthropic or OpenAI, the same as anything you type into a chat. Both run enterprise-grade data protection on their paid tiers. ARCAS never sits in that loop.

Your diagnosis data is stored securely and you can keep it browser-local if you prefer. We do not share data with third parties. For the engagement, the brain runs on your accounts, your tools, your jurisdiction. ARCAS does not host your data on our infrastructure. Your business data stays under your control unless you choose otherwise.

Four groups. Pre-revenue founders who need to find product fit before fixing operations. Companies above fifty people that need a full-time COO, not a fixed engagement. Founders looking for a strategy deck without implementation. And anyone who wants AI to replace their team rather than support it. If any of those is you, the free repo and the playbook are still useful. The paid engagement is not.

Use the free tools. Run Founder OS on yourself. Clone the repo, follow the install, run on it for a month. The paid engagement is sized for founders with revenue and a team that already exists. If you are pre-revenue, the repo is the right level.

The repo and all free tools work for any service business globally. The regulatory content (Emiratisation, e-invoicing, ESR) is UAE and GCC specific. Paid engagements are currently focused on the UAE and GCC because the on-site work is part of how the team-side install lands.

Interior design, recruitment, marketing and PR, facilities management, and other founder-led service businesses. The brain adapts to the industry. The pattern is the same across them: founder holding too much, team waiting for answers, AI subscriptions that have not paid back.

If your business has twenty to forty-nine employees in one of the fourteen targeted sectors, MoHRE Emiratisation mandates apply. Non-compliance fines start at AED 96,000 a year and climb. The diagnosis identifies whether your roles, systems, and onboarding are ready to absorb new hires productively, not just to fill a quota.

Communication complexity jumps. Forty-five potential interactions at ten people becomes one hundred and ninety at twenty. The founder can no longer be in every conversation. At the same time, regulatory obligations move from optional to mandatory. Emiratisation, formalised HR, compliance documentation. The brain holds the context the founder used to hold in their head, so the next ten people don't double the founder's workload.

Beyond Emiratisation, four active pressures. E-invoicing rolling out across 2026 and 2027. Corporate tax at nine percent on profits above AED 375,000, which assumes clean financial systems. Transfer pricing documentation. Economic Substance Regulations audits getting tighter. All four assume you have documented processes someone can audit on demand.

No, and that framing is wrong. The large-organisation playbook of documenting to automate does not apply to a fifty-person business. Here, every person matters. You need systems that make each person more effective, not replaceable. AI comes after people and systems are working, not instead of them.