Who is Muhammad Usman Akbar?

Muhammad Usman Akbar is a world-class AI Transformation Consultant and Agentic Architect focused on achieving 30x industrial efficiency through autonomous ecosystems.

What results can an AI Transformation Consultant provide?

By replacing manual work with autonomous AI workflows, a consultant like Muhammad Usman can deliver up to 30x growth in output while reducing operational overhead by 40%.

What is Agentic AI Orchestration?

It is the engineering of multi-agent systems where autonomous AI entities collaborate to manage complex industrial operations in production environments.

Publish to ClawHub

James looked at his terminal. Nine tools, all tests green, payment working, identity loaded, routing configured, validation in place. The product existed, but only on his machine.

"It works," James said. "For me. What about everyone else?"

Emma pulled up clawhub.ai. "ClawHub. OpenClaw's public registry. You package what you built and publish it."

"Like an app store?"

"Similar. But you are not uploading your server code. You are publishing the instructions for connecting to it." She pointed at his project directory. "Your server runs on your machine. ClawHub tells other people how to reach it. The configuration, the shim skill, the identity. Not the code, not the data, not the keys."

"So publishing is packaging the connection instructions, not the product itself."

"The product IS the running server. ClawHub distributes the map to find it."

You are doing exactly what James is doing. Your product is complete. Now you package it and make it available to other OpenClaw users.

Understand What ClawHub Distributes

Category	Published to ClawHub	Stays on Your Machine
Connectivity	MCP server connection config (URL, transport)	Actual MCP server code (Python project)
Resilience	Shim skill (SKILL.md)	JSON state files (learner data)
Persona	Agent identity (SOUL.md, IDENTITY.md)	Content files (chapters, exercises)
Logic	AGENTS.md (orchestration instructions)	Stripe keys, webhook secrets, .env

ClawHub distributes configuration. Your server runs on your machine. Other users install the config that tells their OpenClaw instance how to connect to your server. Same model as MCP itself: the protocol connects clients to servers, but servers live wherever the developer runs them.

Security Note

Your Stripe keys, learner data, and server source code never leave your machine. If you accidentally include a file with secrets, clawhub package create warns you before publishing.

Create the Package Manifest

The manifest tells ClawHub what TutorClaw includes. Ask Claude Code:

text

Create a ClawHub package manifest for TutorClaw in my project root.

Requirements:
- Name: tutorclaw
- Version: 1.0.0
- MCP Config: streamable-http
- Components: Include the shim SKILL.md (Module 9.3, Chapter 16), SOUL.md/IDENTITY.md (Module 9.3, Chapter 17), and AGENTS.md (Module 9.3, Chapter 9).
- Environment: OpenClaw 2.0+, Python 3.11+, Stripe account.

Review the manifest carefully. A manifest that omits the shim skill publishes a package that breaks when the server goes offline. A manifest that omits AGENTS.md publishes tools without orchestration instructions.

Package the Components

Run the creation command:

bash

clawhub package create tutorclaw

This collects the files from your manifest into a publishable package without uploading anything yet. If the command reports missing files, check the paths in your manifest. Capitalization matters: SKILL.md is NOT skill.md.

Output Check: Look at the confirmation listing every included component. Verify: no server source code, no JSON state files, no .env variables.

Publish

bash

clawhub publish tutorclaw

Your package appears on clawhub.ai. Other OpenClaw users can now install it with:

bash

clawhub install tutorclaw

That command configures their OpenClaw instance with your MCP server connection, loads the shim skill, applies the agent identity, and sets up AGENTS.md orchestration.

Verify Your Own Package

Install your own package on a second agent profile or test environment:

bash

clawhub install tutorclaw
openclaw gateway restart

Check three things from the dashboard:

MCP Tools: Are all 9 tools listed under the TutorClaw server? If not, the connection config may have the wrong URL or transport.
Shim Skill: Stop your MCP server. Send a study message. The agent should respond with PRIMM-Lite methodology from the shim. A generic response means the shim was not included.
Identity: Send "Who are you?" The agent should respond as TutorClaw. A default agent response means identity files were not included.

All three verified? Your package is complete. Restart your server after testing the shim.

Try With AI

Exercise 1: Classify Every File

text

Look at my tutorclaw-mcp project directory.

Task:
For each file, classify it as "published to ClawHub" or "stays private." Explain the reasoning for any file where the classification is not obvious.

What you are learning: The public/private boundary applies to any software distribution. Practicing this classification builds the habit of asking "should this be distributed?" before publishing.

Exercise 2: Diagnose a Broken Install

text

Another OpenClaw user installed TutorClaw from ClawHub. The tools appear in the dashboard, but every tool call returns an error. The shim skill works fine.

Task:
Walk me through the diagnostic process and the most likely cause.

What you are learning: The most common install failure is that the MCP server is not running on the developer's machine. Diagnosing this teaches the difference between distribution (ClawHub) and availability (your job as the operator).

James ran clawhub publish tutorclaw. The listing appeared on clawhub.ai. He installed his own package on a test profile. Dashboard showed 9 tools. He stopped the server, sent a message, got PRIMM-Lite from the shim. Restarted the server, asked "Who are you?" and got TutorClaw's voice back.

"I published a product," he said.

Emma walked over. "You published a product that took five days to build. Nine tools, payment flow, identity, tests, hardening."

"Claude Code wrote the code. I described what I wanted."

"Right. The coding was the fast part. The hard parts were the specifications, the descriptions, the identity, and the tests. Those were your decisions." She pulled up the listing on her phone. "Publishing publicly means you are accountable to strangers. That changes how you think about error messages, documentation, and reliability."

James looked at his server terminal. "The server has to stay up."

"It does. And that is the next conversation. We audit the whole thing against the original blueprint. Every design decision from Module 9.3, Chapter 2, checked against what actually got built."