Personal Brain Agent: Current Status and Future Plans
Table of Contents
I’m excited to introduce the Personal Brain Agent, a new AI plugin for Obsidian.md that seamlessly integrates a powerful agent into your second brain. While AI agents are a well-tested concept among software engineers, my goal with this plugin is to make this powerful technology accessible to a broader, non-technical audience. This post is a deep dive into its features, demonstrating how anyone can transform their workflow by moving beyond simple chatbots.
The Personal Brain Agent is the culmination of a journey I’ve documented in previous articles. It began with an exploration of open standards in “From MCP to ACP: A New Standard in AI Agent Interoperability” and continued with the practical insights from building the initial prototype, which I shared in “Personal Brain: My Journey with ACP in Obsidian”. Now, it’s time to see the results.
Why build on GeminiCLI and ACP? A Focus on Simplicity and Accessibility #
My primary concern when developing this plugin was ease of use for a non-technical audience. The choice of technology reflects this:
- Simple and Cost-Effective Access: By building on GeminiCLI, you authorize access with your existing Google Account. There are no complex API keys to manage or individual tokens to pay for, as access is often included in your current subscriptions. It’s designed to be simple to set up and use.
- Leverage Existing & Approved Tools: Many companies have already introduced tools like GeminiCLI and ClaudeCode for their software engineering teams. This means the Personal Brain Agent uses a tool that may already be approved and available within your organization. Instead of requesting a new application, you are simply using an existing one in a new way, which can make getting access much easier.
- Enterprise-Grade Data Governance: This ties directly into data security. When you use a Google Workspace account, GeminiCLI automatically respects your company’s data governance policies. Your data remains secure and compliant, which is why these tools get approved for professional use in the first place.
- An Open, Interoperable Future: Finally, ACP is an open standard. While the plugin currently uses GeminiCLI, this approach ensures you won’t be locked into a single provider. It opens the door for you to select from a variety of specialized agents in the future.
Chat with Your Agent #
The first thing you’ll notice after installing the Personal Brain Agent is the chat view. This is the main interface where you can begin working with your AI agent.
This view is not just a reskinned AI chat. It’s a fully-fledged AI agent. You can see that in addition to standard responses, the agent accessed a file from your vault to get more knowledge. This is what differentiates an AI agent from a standard AI chat. Instead of manually copy-pasting file contents, the agent can independently access information to assist you.
The Personal Brain Agent is designed for transparency. It reports all actions and its thinking process, and each “Thinking” and “Action” block can be expanded with a simple click. This provides additional context on the agent’s chain of thought and the reasoning behind its actions. You can also see the answer as it’s being generated because the agent supports response streaming. You can easily track its status via the icon on the send button and a spinner at the bottom of the chat, so you’re never left wondering if it’s still working.
Extending the AI Agent’s Context #
An AI agent is only as good as the context it has. Instead of manually pasting content, the Personal Brain Agent allows you to attach any file from your vault, including PDFs and images. This process is made simple with an intuitive auto-complete dialog.
Simply type @ or [[ and start typing a file name. The dialog intelligently sorts files, prioritizing currently open files, then recently opened ones, followed by all other files in your vault.
Attached files are clearly marked, so you can be certain they are being used as context, not just as plain text in your prompt.
As mentioned, you are not limited to Markdown files. Need to analyze a lengthy PDF report? No problem. Attach it to your message, and the agent will extract the necessary information. You can then discuss the contents and even ask it to generate summaries, diagrams, or charts based on the document.
Tools and Permissions #
The Personal Brain Agent comes equipped with a range of tools to act on your prompts. These tools are what allow the agent to become a true assistant. For example, it can follow a link from your notes to fetch the content of a website or search the internet for additional details to help you write an article. Its capabilities include file processing (reading, writing, and searching), fetching web content, and performing web searches.
To ensure you remain in control, any action that could modify your files requires your explicit approval. You will be prompted to confirm before any changes are made.
Future Plans #
Everything presented above is already functional in the Personal Brain Agent, but there’s much more on the horizon. Here are a few items from my backlog:
- Add authentication right from Obsidian. Currently, you need to start Gemini-CLI in the terminal for the first time to log in to your Google Account.
- Easy start for non-technical users (streamlined onboarding).
- Save conversation as a note and restore sessions.
- Integrate user-configurable MCP Servers for custom tools.
- Support for more ACP-compliant agents (e.g., Claude Code).
- Transcribe from recordings.
- Support for slash commands.
- Diff view for tool changes.
- Agent planning visualization.
- For advanced users: an integrated terminal.
- and more…
I’m already using this plugin in my daily work, which helps me identify areas for immediate improvement. I also want to remind you that a closed beta program will be starting soon. I invite you to register and test the Personal Brain Agent for yourself.