The promise of AI was that it would be a true collaborator. But in reality, it's usually just you shoving the same PDFs down its throat every single time you start a new conversation. What it learned yesterday is completely forgotten today. It's incredibly frustrating because it's a machine—it should know.
But it wasn't connected yet.
ContextNest is how you connect every document, interaction, and conversation into something that builds value, scales knowledge, and greatly improves your ability to use AI in the big-picture world your human form operates in.
Here is the exact, step-by-step conversation of how I set up my own ContextNest vault locally in less than 10 minutes alongside my AI engineering partner, and changed the way I work forever.
Step 1: Installing the CLI and Initializing the Vault
Stacey: "...there are now instructions for installing the context nest on the web page https://promptowl.ai/contextnest/. please install for me locally"
Antigravity:
I went ahead and globally installed the open-source CLI using npm, then created a fresh, empty workspace for you to experiment with.
npm install -g contextnest-cli mkdir -p my-knowledge-base && cd my-knowledge-base ctx init --name "My Knowledge Base" ctx verify
Your new local ContextNest vault is ready! All the basic Node setup handles itself perfectly behind the scenes.
Step 2: Choosing an Interface (I chose Obsidian)
Stacey: "does it have a UI?"
Antigravity:
The open-source contextnest-cli we just installed doesn't have its own standalone graphical UI application out of the box.
However, because ContextNest was explicitly built from the ground up to be "Markdown-Native and Obsidian-Compatible," you effectively get to choose your own UI! You can open the folder inside Obsidian, VS Code, Cursor, or any other Markdown editor, and it will work perfectly as your front-end interface. You can type, create wiki links, and tag things normally, while the ctx CLI running in your terminal acts as the "backend" to verify integrity and manage versions.
Stacey: "do i need to download obsidian"
Antigravity:
No, you absolutely do not need to download Obsidian! Because ContextNest is built entirely on open web standards, any code or text editor will work.
However, Obsidian is highly recommended for one specific reason: it comes with out-of-the-box native support for following [[wiki links]] backward and forward, visualizing the knowledge graph, and rendering nested tags (#like/this). It automatically builds the "structural memory" visuals that ContextNest is based on.
Stacey: "lets do obsidian - i don't have any of them"
Antigravity:
I've just successfully installed Obsidian on your Mac for you!
brew install --cask obsidian
To get started with your new ContextNest setup:
- In the Obsidian launch window, choose "Open folder as vault".
- Navigate to your new folder:
my-knowledge-baseand "Open".
Obsidian will instantly recognize the folder, and you can start creating Markdown files, adding tags, linking them together with [[brackets]], and letting ContextNest verify it underneath!
Step 3: Feeding ContextNest with "Source Nodes"
Stacey: "what about adding my emails and slack messages? and transcripts from meetings? I'd like it to work with my google drive."
Antigravity:
According to the ContextNest whitepaper, you don't actually have to manually copy and paste thousands of emails, Slack messages, or meeting transcripts into your folder!
ContextNest handles live, external data through a very specific feature called Source Nodes.
Instead of dumping raw chat logs into your shiny new folder, you create a lightweight Markdown file (the "Source Node") that serves as a bridge. This file contains instructions telling ContextNest how to securely connect to your live Slack workspace, email server, or meeting transcription tool using MCP (Model Context Protocol) or a standard API.
Stacey: "I want transcripts from google drive - they are stored there automatically."
Antigravity:
To get Google Drive transcripts specifically, you would use Anthropic's official open-source Google Drive MCP Server, and connect it directly into your ContextNest "Source Nodes". The beauty of the format is that any MCP Server fits directly into ContextNest using standard syntax!
If you want to pull live transcripts directly from your Google Drive into your local ContextNest vault, here's what you would do:
- Install Anthropic's Google Drive MCP Server
- Create a new markdown file in your Local ContextNest (e.g.
nodes/latest-transcript.md) - Inside the Markdown file, you don't paste the text. Instead, you write a Source Node directive letting ContextNest know it needs to open the Google Drive MCP integration to read the transcript when the AI asks for it.
Step 4: Hooking the Brain to Claude Desktop
Stacey: "yes, let's connect claude and gemini..."
Antigravity:
I just went ahead and modified your claude_desktop_config.json file behind the scenes! Claude Desktop is now officially configured to boot up with two MCP servers actively attached to it:
contextnest(Which points right at your local folder)google-drive(Anthropic's official integration for reading files out of a Google account)
To make your Google Drive transcripts an officially governed part of your ContextNest organization, I just created your very first Source Node document for you.
There is absolutely no transcript text actually saved in that file. Instead, it just contains agent instructions! Whenever you ask Claude a question about a recent meeting, it will see that Source Node in the Context graph, read the instructions, and use the Google Drive MCP integration to dynamically search for and read the live transcript straight from your cloud account.
Step 5: Establishing Structural Memory & Human Governance
Stacey: "but are you recording information to context nest?"
Antigravity:
I am not automatically recording our conversations into your ContextNest vault behind your back.
According to the ContextNest whitepaper, the whole problem with enterprise AI right now is that there's a "vacuum of accountability"—agents automatically dumping or retrieving data without any human approval or versioning.
If I simply started writing raw documents into your folder on my own, it would violate the principle of "Hierarchical Stewardship." I will only ever create new Markdown nodes or record information into your official ContextNest vault when you explicitly ask me to.
Stacey: "Yes, I want you to record everything. I want to use it to generate my weekly reports, and to strategize for PromptOwl, as well as keep track of my sales efforts."
Antigravity:
I love this approach. This fundamentally turns your ContextNest vault into your own personal "Digital Twin"—giving AI agents a perfectly structured, human-governed memory of everything you do!
I just created your first official, governed document: nodes/stacey-weekly-log.md.
I recorded everything we accomplished over the last few days inside it. Whenever we finish a coding session or have a strategic discussion, I will make sure to update your ContextNest log. That way, on Friday, you can simply ask Claude:
"Read my stacey-weekly-log.md and generate a formatted weekly report for the leadership team."
How I Use This Every Day
Setting this up didn't just give me a weekly report generator. It created a "Digital Twin" of my entire workflow. Here's how I use it:
- Strategic Playbooks That Evolve: I created a
sales-playbook.mdand acompetitive-analysis.md. When I'm writing outbound emails, my AI agent references these specific, governed nodes to position us perfectly. But better yet, as I learn new objections from prospects or adapt our positioning in the real world, the AI updates the playbook itself. The playbook is no longer a static PDF that goes out of date the day I write it—it evolves and gets mathematically versioned in perfect sync with my actual real-world work. - Master Task Lists: I have a
task-master.mdnode with actual checkboxes. Because it's open text, I can tell Claude, "Check off the 'Align messaging' task in my master task list," and it literally updates the file. - Live Syncing: Using the Google Drive "Source Node" we created in Step 3, my AI can seamlessly read live meeting transcripts without me ever having to download them.
Everything is connected, governed, and mathematically verifiable.
From Solo to Enterprise: The PromptOwl Alpha
Everything I just described is how you use ContextNest by yourself as an individual. Setting up a locally-governed structural memory for your personal AI is completely free and open-source. You can get started right now:
- Install the CLI via Node:
npm install -g contextnest-cli - Initialize your vault:
ctx init --name "My Workspace" - Learn more about the open-source specification in the ContextNest Whitepaper.
Set it up. Play with it. If you love how this transforms your own personal workflow, you are going to LOVE it for your entire organization.
Scaling to the Enterprise
While the core ContextNest format and CLI are open-source and run beautifully for individuals on a local machine, PromptOwl takes this individual working model and scales it to a governed enterprise level.
If the open-source CLI is Git, PromptOwl is GitHub.
PromptOwl provides the enterprise UI, team-wide hierarchical stewardship permissions, visual relationship graphs, and managed AI injection pipelines that allow entire organizations to collaborate safely on the knowledge they feed to their AI agents. It ensures that your developers are only coding against architecture specs that the CTO has actually approved.
We are currently opening up access to teams who want to move beyond dumping raw files into unverified vector databases.
Apply for Alpha Access
Enter your details below to request early access to the PromptOwl ContextNest enterprise governance layer.