Glossary
Core PuppyOne concepts and terminology.
Data hierarchy
Organization
└── Project
└── Content Node (Mut Merkle tree)
├── folder
├── json
├── markdown
└── fileCore terms
Organization
An organization is the top-level container in PuppyOne. It represents a team or company.
- Manages members and roles, Owner, Admin, and Member
- Manages billing and quotas
- One user can belong to multiple organizations
Project
A project is a data namespace inside an organization, used to isolate data for different business scenarios.
- Each business scenario can have its own project, such as Product Knowledge Base, Support FAQ, or Engineering Docs
- A project contains the content tree, access points, agents, and related resources
- Supports project-level member management and access control
Content Node
Content Nodes are the basic data unit in PuppyOne's cloud file system, organized as a tree and identified by path. Content is stored in a Mut Merkle tree backed by S3.
| Type | Description | Typical use |
|---|---|---|
folder | Folder | Organize directory structure |
json | JSON file | Structured data storage |
markdown | Markdown file | Documents and text content |
file | Binary file | Images, PDFs, and other attachments |
Each Content Node automatically tracks version history through Mut commits and supports diffing and rollback.
Access Point
An access point is PuppyOne's unified abstraction for external integrations. All integration types are managed through the access_points table, with provider and direction fields used to distinguish them.
| Access point type | Description | Examples |
|---|---|---|
| Sync | Data-source sync | Notion, GitHub, Gmail |
| Agent | AI agent instance | Support assistant, analytics agent |
| MCP | MCP protocol endpoint | Tools for Cursor or Claude |
| Sandbox | Code sandbox | Docker or E2B isolated execution |
| Filesystem | Local folder sync | Two-way sync with local directories |
MCP (Model Context Protocol)
MCP is an open protocol introduced by Anthropic that defines how AI models interact with external tools and data sources.
- PuppyOne generates standard MCP interfaces for each agent and MCP endpoint
- Supports direct connections from MCP-compatible clients such as Cursor, Claude Desktop, and Windsurf
- Agents use MCP to call bound tools
FLS (File Level Security)
File-level permissions that precisely control which Content Nodes each agent can access.
- Configured via the
scopefield inaccess_points.config, which defines the paths an access point can reach - Supports permission levels such as read-only and read/write
- Follows the principle of least privilege
Tool
A tool is a data operation that an agent can call, bound to specific Content Nodes.
| Tool type | Description |
|---|---|
query | Query data with filters |
get_all | Read all data |
create | Create a new item |
update | Modify data |
delete | Delete data |
search | Vector search, semantic retrieval |
Access Key
An access key is a credential used for machine and agent authentication, typically in the form cli_xxx.
- Each agent, MCP endpoint, and sandbox endpoint has its own independent access key
- Keys can be regenerated, and old keys become invalid immediately
- Used for non-interactive authentication in the CLI and API
Sync-related terms
Sync connector
A sync connector pulls data from external SaaS platforms into PuppyOne.
- Supports more than 15 platforms, including Notion, GitHub, Gmail, and Google Drive
- Synced data is transformed into structured JSON or Markdown Content Nodes
- Supports incremental sync, scheduled sync, and manual refresh
OpenClaw
PuppyOne's two-way local folder sync protocol.
- Keeps local directories and the cloud content tree in real-time sync
- Runs through a background daemon
- CLI commands:
puppyone access upandpuppyone access down
Collaboration-related terms
Mut Commits
The versioning mechanism for all content in PuppyOne.
- Every write operation creates a new Mut commit that captures the full tree state
- Supports diffing between any two commits
- Rollback restores the tree to a previous commit's state
- Stored in the
mut_commitstable
Audit log
An operation log that records every action performed on Content Nodes.
- Records who, user or agent, did what to which node and when
- Supports filtering by node, time, and actor
- Fully traceable
Distribution methods
MCP server
Provides data access for MCP-compatible clients such as Cursor and Claude Desktop.
REST API
Standard HTTP API for custom agents, backend integrations, and automation scripts. Content operations use the /api/v1/content/{project_id}/... endpoints with path-based addressing.
Publish link
A public short link that lets anyone access JSON data in read-only mode without authentication.
Glossary table
| English | Chinese | Description |
|---|---|---|
| Organization | 组织 | Top-level container for team and billing management |
| Project | 项目 | Data namespace |
| Content Node | 内容节点 | Basic data unit in the cloud file system, identified by path |
| Access Point | 访问点 | Unified abstraction for external integrations |
| MCP | Model Context Protocol | Open protocol for AI model and tool interaction |
| FLS | File Level Security | File-level permissions |
| Tool | 工具 | Data operation callable by an agent |
| Access Key | 访问凭证 | Machine and agent auth key |
| OpenClaw | 本地同步协议 | Local folder two-way sync protocol |
| Sync Connector | 同步连接器 | Data-source sync adapter |
| Mut Commit | 版本提交 | Versioned snapshot of the content tree |
| Audit Log | 审计日志 | Operation tracking record |
| ETL | 数据处理流程 | Extract, transform, load |
| Publish Link | 公开链接 | Public read-only access link |