puppyone-ai/puppyone
Context drive for your AI agents
What it solves
Puppyone provides a centralized, version-controlled "Context Drive" for AI agents. It solves the problem of managing fragmented data across various SaaS tools and ensuring that different agents have secure, scoped access to only the specific files and data they need to function effectively.
How it works
The system transforms data from various sources (GitHub, Gmail, Google Drive, etc.) into agent-friendly formats like Markdown and JSON, storing them in a cloud file system. This drive is then exposed to agents through multiple "Access Points," including the Model Context Protocol (MCP), REST APIs, a CLI, and isolated sandboxes (Docker/E2B) where authorized files are mounted.
Who it’s for
Developers building AI agents and teams managing multi-agent systems that require secure data orchestration, versioning, and audit logs for the context they provide to LLMs.
Highlights
- File Level Security (FLS): Enforces per-agent permissions at the filesystem layer, ensuring agents only see authorized files.
- Git-native Versioning: Includes file-level version history, diff comparisons, and one-click rollbacks.
- Multi-Source Connectivity: Built-in connectors for Google Workspace, GitHub, Supabase, and web pages.
- Diverse Access Methods: Supports MCP for tools like Cursor and Claude Desktop, as well as bidirectional local folder sync.
- Audit Logs: Records every read and write operation for full transparency of agent actions.