In the world of 2026, AI isn’t just a chatbot; it’s an orchestrator. If you’ve been using Obsidian as a static “Second Brain,” it’s time to give it a nervous system.
Oh My OpenCode (and its lightweight sibling, Oh My OpenCode Slim) is a powerful plugin suite for the OpenCode framework. It allows you to run multiple specialized AI agents (the Researcher, the Fixer, the Librarian) directly against your local files. When paired with Obsidian, your vault transforms from a pile of Markdown files into an active workspace where AI agents can read, write, and organize your knowledge in real-time.
🛠️ Prerequisites
Before we dive in, ensure you have the following installed:
- Obsidian: The latest desktop version.
- OpenCode Server: The backend that runs the agents.
- An LLM Provider: Claude (via API), Gemini, or a local provider like Ollama.
🚀 Part 1: Installing the “OpenCode Obsidian” Plugin
The bridge between your vault and the agents is the OpenCode Obsidian plugin.
- Open Obsidian Settings > Community Plugins.
- Browse for OpenCode Obsidian (by windyboy) and click Install then Enable.
- Configure Permissions: This is crucial. Under the plugin settings, set your Tool Permission Level:
- read-only: Agents can only read your notes.
- scoped-write: Agents can write to specific folders (recommended for safety).
- full-write: Agents can edit anything in your vault.
🧩 Part 2: Setting Up Oh My OpenCode (Multi-Agent Power)
While the base plugin gives you an assistant, Oh My OpenCode adds the “Professional” agent stack. It installs a suite of specialized sub-agents that work in parallel.
1. Installation
Oh My OpenCode is typically installed via the OpenCode CLI or by placing the agent files in your vault’s hidden configuration folder.
- Navigate to your vault’s root in your terminal.
- Create the .opencode directory if it doesn’t exist.
- Clone the Oh My OpenCode agents into .opencode/agent/.
2. The Agent Hierarchy
Once installed, you’ll have access to:
- @Librarian: Best for cross-linking notes and maintaining your MOCs (Maps of Content).
- @Researcher: Automatically browses the web and summarizes findings into new .md files.
- @Oracle: Answers complex questions based strictly on your existing vault data.
📖 Tutorial: Your First Autonomous Workflow
Let’s set up a “Research-to-Knowledge” loop.
Step 1: Initialize a Project
Create a new note in Obsidian titled Project Alpha. In the OpenCode chat sidebar (which appears after installing the plugin), type:
/command @Researcher find the latest benchmarks for the Qwen3-Max model and save them to a new note in my /Research folder.
Step 2: The Multi-Agent Handover
Once the Researcher finishes, the OpenCode Server keeps the context. Now, call the Librarian:
@Librarian review the new note in /Research and link it to my ‘AI Models’ Map of Content. Add a summary to the ‘AI Models’ note.
[Image showing a flow diagram of the Researcher agent creating a file and the Librarian agent updating an existing index file in Obsidian]
Step 3: Verify the “Green Dot”
Look at the bottom of your Obsidian chat sidebar. A Green Dot indicates a live connection to the OpenCode Server. If it’s red, ensure your OpenCode backend is running in the background.
💡 Pro-Tips for Power Users
| Feature | Best Practice |
| Session Management | Use the New Chat button frequently. Long sessions confuse agents; start fresh for every new task. |
| Safety First | Use .opencode/config.json to “deny” the AI access to sensitive folders like Private/ or Finances/. |
| Custom Skills | You can write your own “Skills” in Markdown. Place them in .opencode/skills/ to teach your AI how to use specific Obsidian plugins (like Dataview). |
🏁 Conclusion
The combination of Oh My OpenCode and Obsidian moves us away from “Chat-with-PDF” and toward a truly Agentic Vault. You aren’t just writing notes anymore; you’re managing a team of digital interns who live inside your markdown files.
Ready to try it? Start by moving one project into a scoped folder and letting the @Librarian organize your messy thoughts!