Distribution Overview

Put your Context File System in the hands of agents.

What distribution means

You have already gathered data from many sources into PuppyOne through Connect, and organized it through Data Management. The next step is to make that data available to agents. That is distribution.

PuppyOne offers multiple distribution methods, covering everything from IDE integrations to fully custom agents. In every case, the goal is the same: let agents read and write your Content Node tree.

Distribution methods at a glance

Which one should I use?

What kind of agent are you using?
│
├─ Cursor IDE
│  └─► Use Cursor MCP integration (the easiest path)
│
├─ Claude Desktop
│  └─► Use Claude Desktop MCP configuration
│
├─ Claude Code
│  └─► Use local file sync (Claude Code is file-first)
│
├─ Other MCP-compatible clients (Cline / Continue / Zed, etc.)
│  └─► Use custom MCP client configuration
│
├─ A custom agent / script
│  └─► Do you need to execute code in a sandbox?
│     ├─ Yes → Use Code Sandbox
│     └─ No → Use REST API
│
└─ Not sure
   └─► Start with Cursor or Claude Desktop and get running in 5 minutes

Core concepts

MCP (Model Context Protocol)

MCP is a standard protocol that lets AI models access external tools and data. PuppyOne generates an MCP Server URL for each connection, and any MCP-compatible client can connect to it directly.

Your MCP Server URL looks like this:

https://api.puppyone.ai/api/v1/mcp/server/{API_KEY}

File sync (OpenClaw)

For file-first agents such as Claude Code, PuppyOne uses the OpenClaw protocol to establish real-time two-way sync between the cloud and a local folder. When the agent reads or writes local files, it is effectively operating on your Context File System.

REST API

If you are building your own agent or automation flow, you can call PuppyOne's HTTP API directly to operate on Content Nodes.

Code sandbox

PuppyOne can mount your data into an isolated sandbox environment, either Docker or E2B microVMs, so agents can execute code safely against your data.

Quick start

No matter which method you choose, start with these steps:

1. Create a project and import data. See Quickstart
2. Create a connection. Choose a distribution method in the dashboard, or create one with the CLI

# Install the CLI
npm install -g puppyone
 
# Log in
puppyone auth login
 
# Create an MCP connection (for Cursor / Claude Desktop / custom clients)
puppyone conn add mcp "My Data"
 
# Create a local folder sync (for Claude Code)
puppyone conn add folder ~/project/context --name "Claude Code Workspace"
 
# Create a sandbox (for code execution)
puppyone conn add sandbox "Python Runner" --type e2b

See the corresponding page for detailed setup instructions for each distribution method.