Bring Your Own Claw
What's Your OpenClaw Strategy?
Last week Jensen Huang asked a question that caught a lot of people off guard: “What's your OpenClaw strategy?” Most companies are still figuring out their AI strategy. Now they need a strategy for the strategy.
The Reality
Here is what seems fairly certain: people will set up their own OpenClaw instances at home. They will customize them, teach them their preferences, train them on their personal workflows. These agents will know how they think, how they write, what they care about.
Then Monday morning comes, and they walk into an office where they either don't have access to their personal agent — or aren't allowed to use it.
Sound familiar? It should.
Øyvind Brekkhus Sandåker
Co-Founder & Chief AI Officer
Day of Week
We Have Seen This Movie Before
Remember when Facebook launched Groups? Almost overnight, entire companies organized their internal communications into Facebook groups. Project updates, shift schedules, customer issues — all flowing through a platform the company didn't own, couldn't audit, and had zero control over.
It felt convenient at the time. In hindsight, it was a security and compliance nightmare. Sensitive customer data in public groups. Employee disputes documented on a social media platform. No off-boarding process — when someone left the company, they still had access to everything.
OpenClaw is the same inflection point, but the stakes are higher. A personal AI agent doesn't just read your company data — it acts on it.
OpenClaw is the Facebook Groups problem all over again — except this time the tool doesn't just read your data. It acts on it.
The Two Camps
Some organizations will say: “You can't bring your personal claw here. Use ours.” Perfectly reasonable. Regulated industries, sensitive data, compliance requirements — there are plenty of good reasons.
Others will say: “Plug your own claw into our systems. Let's see what happens.” Also reasonable. Faster onboarding, happier employees, more innovation.
But here is the uncomfortable truth about the first camp: when you tell an employee to use the company's claw instead of their own, it is like handing them a new laptop with no bookmarks, no shortcuts, no history. Their personal OpenClaw knows how they structure their thoughts, what tone they prefer, which workflows they have refined over months. The company claw knows nothing.
That gap matters. And if we don't solve it, people will just quietly use their personal claw anyway — exactly like they did with Facebook Groups.
Camp A
“Use our claw”
- Full control over data and access
- Compliant with security policies
- Consistent tooling across the org
- Risk: feels limiting, employees may bypass it
Camp B
“Bring your own claw”
- Employees already know their agent
- Faster onboarding, less friction
- More innovation from personal workflows
- Risk: shadow AI, data leaks, no audit trail
The Bridge: Portable Agent Memory
The real solution is neither camp in isolation. We need something in between: a way for an enterprise claw to bootstrap itself from what the personal claw already knows.
Think of it as portable agent memory. Preferences, communication style, domain knowledge, workflow patterns — things that are about you, not about the company's proprietary data. When you join a new org or switch to a company-managed claw, it should be able to import the parts that are yours to carry.
Not everything, obviously. Access credentials stay behind. Company-specific context doesn't transfer. But the things that make the agent feel like your agent? Those should travel with you.
We haven't fully solved this yet. Nobody has. But it's the problem that will determine whether “use our claw” feels like a downgrade or a natural continuation.
“Changes don't happen automatically. You are the human in the middle. Not an admin. Not the system. You.”
Øyvind Brekkhus Sandåker — Co-Founder & CAIO, Day of Week
How We Are Answering This at Day of Week
We are built around AI, people, and local stories. The stories — the producers, the places, the people behind the food — are the most important data in our system.
For us, “OpenClaw as a Service” means your agent needs to be able to talk to our systems. Not by screen-scraping a dashboard. Not by pretending to be a browser. Through a proper interface.
We landed on:
- An open-source CLI (@dayofweek/dcli on npm) — a deliberately thin layer over the platform
- A Day of Week Skill that any OpenClaw instance can install (apparently MCPs are out..?)
- Designed to plug straight into your own agent, wherever it runs
Your claw can read all data you have access to, propose changes to the data model, perform actions in the system, and send orders to producers. But — and this is important — nothing happens automatically. Proposals land in Day of Week Designer, where you approve what actually gets executed.
But don't take my word for it. Here's one of my claws explaining it better than I can:
Tropheus, one of my OpenClaw instances, explaining the Day of Week Skill
What This Looks Like: A Saturday Drive
You are in the car with the family, heading up the coast. The kids are arguing about what to have for dinner. You remember that farm you visited last summer — the one with the incredible lamb.
You tell your claw: “Check if Øvre Ådal Gård has anything available this weekend. If they do, place an order for Saturday pickup. And find somewhere to stop for coffee on the way.”
Your claw spins off the tasks. It checks the farm's inventory through the Day of Week platform, finds that they have fresh lamb available, drafts an order, and sends you a confirmation to approve. Meanwhile, it cross-references your route with local cafes and suggests a stop at a place you haven't tried before — one that sources from producers in the same network.
You glance at the notification, tap approve, and go back to settling the dinner debate. (The lamb wins.)
The system isn't just information-driven. It's action-driven.
Oh, and MCP? Apparently That's Over Now.
If you spent the last six months implementing MCP integrations, I have some news from the oracle that is Silicon Valley: MCP is out. Skills are in.
I say this with affection for everyone who just got their MCP setup working. The valley moves fast, and it has the attention span of a caffeinated goldfish. Last quarter MCP was the future. This quarter it is legacy technology, apparently.
The shift toward Skills — modular, composable, user-owned capability bundles — actually makes sense though. Skills are portable. They travel with the agent, not with the server. They can be installed, updated, and shared independently. For our use case, this is exactly right: you install the Day of Week skill into your claw, and it immediately knows how to interact with our platform.
So yes, we built on Skills. Not because Silicon Valley told us to (they hadn't decided yet when we started), but because it was the right architecture for letting any agent connect to a real system without requiring the system to expose itself as a giant API surface.
Silicon Valley has the attention span of a caffeinated goldfish. Last quarter MCP was the future. This quarter it's legacy technology.
Do We Still Need Interfaces in 2026?
A question I keep coming back to. For the end user — the person on the road trip, placing an order through their claw — maybe the traditional UI matters less. The agent is the interface.
But for the people who own the systems that these agents connect to? The need for rich, well-designed interfaces actually increases. When customers and employees start using your data and systems in entirely new ways — ways you never designed for — you need better tools for control, insight, and approval. Not fewer.
The agents handle the doing. The interfaces handle the deciding.
What your claw can do
Via the Day of Week CLI & Skill
- Read all data you have access to
- Propose changes to the data model
- Perform actions in the system
- Send orders to producers
- All with your approval in the loop
So, What's Your Strategy?
At Day of Week we support both camps. If you want a ready-made claw that works with our platform, we have that. If you want to bring your own claw and connect it via our open-source CLI, we have that too.
The question is not whether your employees will use AI agents. They already do. The question is whether you will give them a sanctioned path — or wait until they find their own.
Define your policy
Decide where personal agents are welcome and where they aren't. Make it explicit.
Provide alternatives
If you restrict personal claws, offer a company-managed one that doesn't feel like a downgrade.
Build the bridge
Enable portable preferences so switching from personal to company claw isn't starting over.
What's your OpenClaw strategy?
Whether you're exploring how to let agents interact with your systems, or figuring out your BYOC policy, we would love to hear from you. We are actively building this and looking for partners who want to shape it with us.
We support both
- Ready-made claw for our platform
- Bring Your Own Claw via open-source CLI
- Day of Week Skill for any OpenClaw instance
- Human-in-the-loop approval for all actions