I Built an AI Employee That ACTUALLY Works For Me (Opus 4.5 + Obsidian)
A comprehensive tutorial showing how to integrate Claude Opus 4.5 with Obsidian to create an AI agent workspace that can autonomously handle diverse tasks like content creation, research, and web searches.
Published Dec 9, 2025 by Riley Brown
Key Insights
Claude Opus 4.5 (via Claude Code) can function as a powerful general-purpose agent beyond just coding when properly set up with a structured workspace.
Integrating Claude Code with Obsidian creates a persistent, organized knowledge base that the AI can maintain and reference for any task.
The README file serves as the central instruction manual for the AI agent, providing context and organizational guidance for every interaction.
The system can perform complex tasks like web research, content creation in specific styles, and strategic planning without constant human intervention.
Creating examples folders with high-quality reference content significantly improves the AI's ability to generate contextually appropriate outputs.
Folder organization is crucial for maintaining a scalable workspace that can grow to handle numerous projects and reference materials.
The future potential involves integrating more specialized tools (like image generation) to further expand the AI's capabilities within this framework.
0:00
Intro
“Claude Opus 4.5 inside Claude Code is the most powerful coding agent in the entire world. But it's not just used for coding. It's actually the most powerful general-purpose agent in the entire world.”
Riley Brown introduces the core concept of the video: using Claude Opus 4.5 through Claude Code as a general-purpose AI agent rather than just for coding. He explains that when integrated with Obsidian (a note-taking app), users can create a workspace that Claude Code will organize and maintain independently.
The introduction outlines the capabilities of this setup, including web searching, content research, scripting, planning, and even image generation with the right tools. Brown positions this as a way to have an AI 'co-CEO' or employee that can handle full tasks with delegation, promising viewers they'll learn to set up this system by the end of the video.
Takeaways
Claude Opus 4.5 can function beyond coding as a comprehensive general-purpose agent
Integration with Obsidian creates an organized, maintainable workspace
The system enables delegation of complete tasks to the AI
Potential capabilities include web searches, research, scripting, and planning
When properly set up, the AI can function similar to a digital employee
1:04
Claude Code, Cursor, Obsidian Setup
“We're going to be using Obsidian, which is where all of the files are going to be located and viewable and editable in a nice clean note-taking app.”
This section covers the technical setup required to create the AI agent workspace. Brown explains the three main tools used in the demonstration: Obsidian (for file organization and note-taking), Cursor (his preferred interface for using the Claude Code extension), and Claude Code itself. He notes that VS Code can be used as a free alternative to Cursor.
The setup process begins with creating a new vault in Obsidian named 'agent workspace' and establishing a README file, which Brown emphasizes as the most critical component of the system. He demonstrates how to connect Cursor to the workspace, install the Claude Code extension, and begin the integration between the tools.
Takeaways
The system requires three tools: Obsidian (for notes), Cursor/VS Code (for interface), and Claude Code extension
VS Code provides a free alternative to Cursor for those who prefer it
Creating an Obsidian vault serves as the foundation for the agent workspace
The README file is identified as the most important component in the workspace
Integration between Obsidian and Claude Code creates a seamless workflow
5:18
Creating README for the Agent
“This readme is going to be the thing that your agent Claude Code will read every time you give it a task. And for this reason, it is the most important file in your entire notebook.”
Brown focuses on creating the crucial README file that serves as the instruction manual for Claude Code. He demonstrates how to instruct the AI to create this file, explaining that it's what the agent reads every time it receives a task, making it the central reference point for all operations.
The README is structured to include information about the workspace purpose (in this case, marketing for Vibe Code app), organizational instructions for different types of content, and reference materials. Brown shows how Claude Code can autonomously create folders for different content types (ads, content, research, and strategy documents) and maintain this organization system as the workspace grows.
Takeaways
The README file provides essential context and instructions for every AI interaction
It should include the workspace purpose, organization structure, and reference materials
Claude Code can autonomously create and maintain the folder structure
The system becomes more valuable as it scales with more content
Organization instructions in the README help maintain consistency across projects
6:55
Generating Ads based on Competitor Ads Page
“I want these to be the first five ads that we create on an app. I want this to be on the vibe code app.”
This section demonstrates Claude Code's ability to generate marketing content by creating ads for the Vibe Code app. Brown first adds competitor ad scripts to the workspace as reference materials, then instructs Claude Code to clean up and organize these examples. He then directs the AI to create five different types of ads (UGC style, education style, side hustle style, and one targeted at moms).
Claude Code seamlessly executes these tasks by referencing the documentation for the Vibe Code app (accessed via URL), updating the README with product information, and generating high-quality ad scripts organized into appropriate folders. The AI even conducts additional research on effective ad strategies and proposes additional ad concepts beyond what was initially requested.
Takeaways
Claude Code can generate diverse marketing content by referencing competitor materials
The AI accesses and processes web documentation to understand the product
Organization of content into appropriate folders happens automatically
The system can conduct independent research to improve outputs
Generated content maintains consistent formatting preferences specified by the user
16:41
Long Form Content YT Transcript
“Let's go to someone we want to create content like. Let's go Andre Jick. He is a very good content creator and he writes very good scripted content.”
Brown shifts focus to demonstrate how Claude Code can handle long-form content creation. He finds a YouTube video by content creator Andre Jick, extracts the transcript, and provides it to Claude Code as an example of high-quality scripted content. The AI organizes this reference material into an examples folder structure.
Brown shows how the system can identify and remove sponsorship segments from reference material automatically after simple instruction. This highlights the AI's ability to learn preferences and maintain them across tasks without repeated direction, creating a more efficient workflow over time.
Takeaways
Reference materials from successful content creators help the AI understand style preferences
Claude Code can automatically organize example content into appropriate folders
The system learns and implements preferences (like removing sponsorships) across tasks
Having diverse reference materials improves the AI's content creation capabilities
The examples folder structure helps maintain organization as reference materials grow
19:07
Creating long form script
“I want you to create a script in the style of Andre Jick about the Vibe Code app.”
This section showcases Claude Code creating an original long-form video script in Andre Jick's style about the Vibe Code app. Brown instructs the AI to research his YouTube channel for context, review Vibe Code documentation, and then write a script mimicking Andre Jick's presentation style. The AI conducts multiple web searches to gather necessary information before crafting the script.
The resulting script successfully captures Andre Jick's conversational style while accurately representing the Vibe Code app's capabilities. Brown expresses approval of the script, noting it's something he might actually use, demonstrating the practical value of the AI's content creation abilities. The section illustrates how the agent combines style analysis, product research, and creative writing into a cohesive workflow.
Takeaways
Claude Code can analyze and mimic specific content creator styles
The system conducts necessary research before creating content
Web search capabilities allow the AI to gather current, relevant information
Generated content can be production-ready with minimal human editing
Style emulation combined with accurate product representation creates authentic content
21:08
Reviewing the Readme and summarizing the agent process
“You can see here that in this video, we created a readme which gives instructions every time this agent is run. It puts this readme in the context. And it reads the readme first.”
In the final section, Brown reviews the entire system they've built, emphasizing how the README file functions as the central instruction set that provides context for every agent interaction. He explains how the workspace organization has evolved through the demonstration to include content, examples, research, and various organizational elements.
Brown then shares his vision for future developments, suggesting that tools like Nano Banana, VO, and Sora could be integrated with Claude Code to enhance capabilities such as image generation for thumbnails. He predicts that future AI models like Claude Opus 5 will significantly improve context management, enabling workspaces with potentially millions of folders and advanced analytics capabilities. He mentions that his team is already implementing similar workspaces across different departments.
Takeaways
The README file serves as the foundation for all agent operations
Workspace organization evolves organically as the system is used
Future integrations with specialized AI tools will expand capabilities
Next-generation AI models will handle exponentially larger context windows
The approach is already being implemented across different business functions
Conclusion
The integration of Claude Opus 4.5 with Obsidian represents a significant advancement in AI assistance, transforming a coding tool into a comprehensive digital employee capable of handling diverse tasks with minimal human intervention. By creating a structured workspace with clear instructions and examples, users can delegate increasingly complex work to an AI that continuously improves its understanding of preferences and objectives.
This approach demonstrates how current AI technologies can be leveraged beyond their intended purposes to create powerful custom solutions. The key innovation isn't just the AI itself but the framework for interaction—using a README as a persistent memory and instruction set, coupled with organized reference materials that provide context for any task.
So what? This represents a shift from treating AI as a reactive tool to functioning as a proactive agent that maintains its own workspace and learns from each interaction. For professionals facing time constraints, this methodology offers a scalable approach to delegation that works within existing limitations of AI while maximizing its strengths. As AI models continue to advance, these workspaces will only become more capable, potentially revolutionizing how individuals and organizations manage information, conduct research, and create content.