PuppyOne

What is PuppyOne?

PuppyOne is a local-first file workspace where people and AI agents share business context. It syncs Notion, Slack, Gmail, Postgres, and other business apps into plain files, then gives each agent scoped access through MCP, Bash, REST API, or a sandbox. Git-style history tracks every edit and supports instant rollback.

Top Features:

• Scoped access: give each agent its own access point with folder limits.
• Version control: every write becomes a commit with agent identity and diff.
• Connectors: sync over 15 sources like Slack, Jira, and Airtable into files.

Use Cases:

• Multi-agent projects: let research and coding agents share one folder with separate permissions.
• Team context: publish a shared workspace so teammates and agents see identical files.
• Safe experiments: branch before risky agent runs and roll back mistakes instantly.

Who Can Use PuppyOne?

• AI developers: connect Claude Code, Cursor, or CrewAI to one governed file store.
• Operations teams: keep product knowledge and customer data where agents can read it.
• Security teams: self-host the open source version so files stay inside your network.

Pricing

• Free ($0): one member, one project, three agent access points, 2 GB storage.
• Developer ($19 per month): three projects, ten access points, 25 GB, and 180-day history.
• Team ($99 per month): ten members, 100 access points, 250 GB, roles, and one-year history.

Pros and Cons

Pros:

• Open source: the Apache 2.0 code runs on your own servers with Docker.
• Agent agnostic: any file-based agent connects through MCP, Bash, API, or sandbox.
• Clear audit trail: trace every human or agent edit back to its source.

Cons:

• Developer focus: setup assumes comfort with MCP, command lines, and agent tooling.
• Tight free tier: one project and three access points limit larger multi-agent tests.
• Young product: the desktop app and connectors still change quickly between releases.

FAQs:

1) Is it open source?

Yes, it is open source and can be self-hosted with Docker.

2) Which agents can connect?

Claude Code, Cursor, Codex, LangChain, CrewAI, n8n, and any MCP-compatible agent.

3) Can I undo an agent mistake?

Yes, every change is a commit, so you can roll back anytime.

4) Must files live in the cloud?

No, files stay on your computer until you choose to publish them.

5) How is it different from a vector database?

Agents get readable files they can grep and edit, with permissions and history.