Thinking about OpenClaw? Start here

Hey Vibe Marketer,

OpenClaw is one of the most interesting things happening in AI right now.

Because instead of just chatting with AI, you can run agents that do work for you.

Even while you sleep.

A few days ago in a Vibe Marketer’s mentor session, James Pasmantier, founder of NimbleDraft walked through how he built a working OpenClaw system.

In the 90-minute session he showed:

• how his AI advisory board of agents works
• how OpenClaw connects to tools like Notion and Telegram
• and how to run agents safely with the right guardrails

If you're curious about OpenClaw but cautious about installing it, this walkthrough is a great place to start.

So before you spin up your first OpenClaw agent, here’s what you should understand:

• what OpenClaw actually is
• how is it different from Claude Code
• how safely you can spin up agents in OpenClaw

So, what is OpenClaw and how is it different from Claude code?

OpenClaw is an open-source AI assistant system that lets you run AI agents that can perform tasks automatically.

You can message the agent like a personal assistant, and it can research, write, monitor updates, and work in the background.

OpenClaw often gets compared with Claude Code. And, they’re built for very different purposes.

Here’s the easiest way to understand the difference:

So, here’s what to set up before running OpenClaw agents

1. Don’t run agents on your personal laptop: Use a separate machine or server so your personal files and accounts stay protected if something goes wrong.

2. Restrict who can control your agents: Set up an allowlist so only your account can message or control them.

3. Agents draft. You publish: Let agents research, write, and summarize — but keep final publishing in human hands.

4. Don’t connect sensitive systems too early: Avoid giving agents direct access to payments, banking, contracts, or production databases.

5. Know how to stop them: Before you begin, make sure you know how to stop a session, shut down the agent gateway, and set spending limits.

6. Review early outputs closely: Treat your first agents like new hires. Check what they do before trusting them with more.

7. Teach them to challenge you: Tell your agent to push back when your thinking is weak. That makes it more useful for strategy and research.

8. Start with one agent: Don’t build a full multi-agent setup on day one. Start with one, improve it, then expand.

Here’s what the session covered:

One of the most interesting ideas James shared was building an AI advisory board using OpenClaw agents.

Instead of one generic agent, you design agents that think through the lens of specific experts or roles.

1. Agents should have clear roles: Rather than one “do everything” assistant, create agents for specific functions like research, marketing, partnerships, or strategy.

2. Give agents expert perspectives: James maps agents to trusted experts and frameworks so their recommendations reflect different ways of thinking.

3. Keep a central source of truth: The advisory board works best when agents pull context from a single knowledge base (for example Notion).

4. Agents collaborate, not just respond: In his setup, agents can pass work to each other — for example research → strategy → content.

5. Start small, then expand: The system begins with one or two agents solving real problems, then grows into a broader advisory board.

The result is a system where agents don’t just answer prompts they help evaluate ideas, monitor opportunities, and generate work continuously.

Note: If you don’t want to build this from scratch, James has also created an Advisory Board Skill and Community members get 50% off.

Want the full walkthrough and resources?

Inside the vibe marketer’s community, you’ll find the full 90-minute session recording plus the resources James Pasmantier shared, including:

• Agent architecture guide

• OpenClaw Session slides

• How to build your first agent

• Detailed security checklist

Over 3,000 marketers are already experimenting with AI inside the community. If you're exploring this space, come build with us.

—

Miscellaneous vibes

1. Claude Code has a /loop skill that feels underrated

2. How to make slides with Claude…

3. GitHub repo that lets you spin up an ai agency with ai employees

4. Scheduled tasks in Claude Code desktop…

5. How to force your agent to obey your design system

6. Claude Cowork with the right context structure goes from generic assistant to executive-level partner

7. Claude Cowork Masterclass for Beginners

Let’s vibe

Crazy time to be building with AI. And, looks like Claude code is leading the game…

Wanna scale marketing and outperform your competitors? Upgrade your AI.
Wanna learn from marketers actually shipping with AI? Upgrade yourself.

Got feedback or a question? Just hit reply. I read every message.

Know someone who’d love this newsletter? Share it with a friend who is figuring AI out too.

The Vibe Marketer