# Plinth: Let your teams build with AI. On rails you set.

Governed apps, built by anyone, running in your cloud.

Plinth lets teams build apps and dashboards with the AI assistants they already use (Claude, ChatGPT, Cursor, GitHub Copilot) while people review the risk, IT controls deployment and every app runs in your own cloud.

- ✓ Any AI assistant
- ✓ Your cloud
- ✓ People-only approvals

[Book a demo](https://plinth.example/#demo)

## Is Plinth for you?

If your teams build apps and dashboards with AI, let them. With Plinth. Same AI tools. Reviewed by people. Running in your cloud.

1. **Your builders: build with their own AI.** Claude Code, Codex, ChatGPT or Copilot. Plinth keeps every build on your rules. Submit from the chat.
2. **Their managers: approve in plain English.** Each app carries an app passport: what it does, what data it uses, who it is for and what it saves. No code to read.
3. **Your IT team: ship to your cloud.** One click deploys what managers approved. Private, checked, on the record. The deployed image is verified to match the approved code.

Every build passes Plinth checks:

- Approved template
- Approved data source
- Single sign-on
- Private network

Plinth is built with design partners in healthcare, financial services, government and higher education.

## Under the hood: three parts, one governed path to production

Your people govern in the console, their agents build with the SDK, and the apps run in your cloud. AI agents such as Claude, ChatGPT and Copilot (or any MCP client) build through one governed MCP connection.

### Plinth console

For admins, architects and approvers. Where your people capture the estate and govern every app: the register of systems and capabilities, app reviews with the passport beside every decision, and the inventory with cost, audit and a switch to stop any app.

Approvals are routed by risk. For example, an app that reads patient appointments needs:

- Manager: Approved · Teams
- Data owner, because it reads appointments: Approved
- Privacy officer, because it's health data: Approved · fresh MFA
- IT: deploys once every approval is in

Small changes that keep the same data and audience go to the app owner only. Low-risk, read-only team apps take a one-click fast lane.

### Plinth SDK

For your teams' AI agents. What agents build with, in Claude, ChatGPT, Copilot or Cursor. Plinth's MCP server keeps the agent on your guidelines while it works; the SDK makes the enterprise way the easy way inside the app.

Plinth MCP, while the agent builds:

- Asks the estate before writing data code
- Starts from approved templates
- New outside destinations need sign-off
- Can request approval, never grant it

Plinth SDK, inside the app:

- Single sign-on at the door
- Data only through approved interfaces
- Logging and usage measured for you
- No secrets or keys in the code

### Plinth data plane

In your cloud. The governed landing zone where Plinth-managed apps run, in your own cloud subscription, next to your data. Plinth runs the platform; your data never leaves your estate. Supported compute: Azure, AWS, Google Cloud or on-prem. Your identity provider, your firewall, your Git.

- **Private by default.** Your network, your sign-in, private data stores.
- **Only approved destinations.** Apps reach only what Security lists, through your firewall.
- **Stop any app in a minute.** One switch for IT or Security; code and data are kept.
- **Every action on the record.** Tamper-evident audit, streamed to your security tools.

## Security

- Private by default
- People-only approvals: AI agents can request approval, never grant it
- Your identity provider (single sign-on)
- Audit-ready evidence: tamper-evident audit trail, streamed to your security tools
- Data hosted in Australia

## Let your teams build. Keep the keys.

See Plinth on your own estate in 30 minutes: [Book a demo](https://plinth.example/#demo).