Claude Code
Use local file sync so Claude Code can read and write your PuppyOne data directly.
Why file sync instead of MCP
Claude Code is an AI coding tool built around file operations. It works directly with the local file system, and unlike Cursor, its strongest workflow is local file editing.
That is why PuppyOne provides a file sync workflow for Claude Code. The OpenClaw protocol establishes real-time two-way sync between PuppyOne in the cloud and a local folder. Claude Code edits local files, and the changes sync back to the cloud automatically.
How it works
PuppyOne cloud Local folder
┌──────────────────┐ ┌──────────────────────┐
│ /notion-docs/ │ ◄──────────► │ ~/project/context/ │
│ /github-issues/ │ real-time │ /notion-docs/ │
│ /linear-tasks/ │ two-way sync│ /github-issues/ │
└──────────────────┘ └──────────┬───────────┘
│
Claude Code reads and writes here- Cloud -> local: Data imported into PuppyOne through connectors, such as Notion pages or GitHub issues, syncs automatically into the local folder
- Local -> cloud: File changes made by Claude Code are pushed back to PuppyOne automatically
- Conflict handling: If both sides change the same content, PuppyOne detects conflicts and provides resolution options
Prerequisites
- Claude Code, installed and available
- Node.js 18+
- A PuppyOne account with data already imported into your project
Setup
Option 1: Full CLI flow
Step 1: Install the CLI
npm install -g puppyoneStep 2: Log in
puppyone auth loginComplete authorization in your browser when prompted.
Step 3: Select a project
puppyone project ls
puppyone project use "My Project"Step 4: Create a folder connection
puppyone conn add folder ~/project/context --name "Claude Code Workspace"This creates a two-way sync connection in PuppyOne and returns an access key.
Step 5: Start sync
puppyone access up ~/project/contextOnce sync starts, Content Node data from PuppyOne is downloaded into ~/project/context/, and later changes sync in real time.
Option 2: Quick dashboard setup
- Open the PuppyOne dashboard
- Go to Project -> Connections
- Click Add Connection and choose OpenClaw / Local Folder
- Enter a name and create the connection
- Copy the generated access key
- Run this in your terminal:
npm install -g puppyone
puppyone access up --key cli_xxx ~/project/contextReplace cli_xxx with the access key you copied.
Using it in Claude Code
After sync starts, your PuppyOne data is available in the local folder. Open Claude Code and include that sync directory in your project.
Read data
Please read all documents under context/notion-docs/
and prepare a technical architecture overview.Claude Code reads the local Markdown and JSON files directly, with no API calls required.
Modify data
Update the price of Widget Pro in context/products/pricing.json to 149.99.Claude Code edits the local file, and the change syncs back to PuppyOne automatically.
Create new files
Based on the meeting notes in context/meeting-notes/,
generate a weekly report and save it to context/reports/weekly.md.New files are automatically synced into PuppyOne as new Content Nodes.
Sync management
Check sync status
puppyone statusThis shows all active sync connections and their status.
Stop sync
puppyone access down ~/project/contextpuppyone access up handles connection setup, initial sync, and starting the local daemon. When you want to stop it, use puppyone access down.
Directory structure
The synced local folder mirrors the Content Node tree in PuppyOne one-to-one:
~/project/context/
├── notion-docs/
│ ├── Product Requirements.md
│ └── Technical Architecture.md
├── github-issues/
│ ├── issue-123.json
│ └── issue-456.json
├── products/
│ └── pricing.json
└── reports/
└── weekly.md- Folder ->
foldernode type in PuppyOne .jsonfile ->jsonnode type in PuppyOne.mdfile ->markdownnode type in PuppyOne
FAQ
Sync starts but no files appear
- Make sure your PuppyOne project contains data
- Check that the access key is correct
- Run
puppyone statusto inspect connection status
File changes do not sync back to the cloud
- Make sure the sync process is still running
- Large files or many files may take a few seconds
- Make sure the file is inside the sync directory
What happens if there is a conflict?
When the same file is modified both locally and in the cloud, PuppyOne will:
- Preserve both versions
- Keep the pre-overwrite content in version history as a safety net
- Let you inspect diffs in the dashboard and decide whether to roll back
Can I sync only part of my data?
Yes. The safest approach is to create a dedicated folder or workspace in PuppyOne for content you want to expose to Claude Code, then run puppyone access up on the corresponding local directory.
Next steps
- Custom MCP Clients. Use MCP in other AI tools
- REST API. Access PuppyOne programmatically over HTTP