Security tools tend to live in one place, and the work happens somewhere else. You read a finding in a dashboard, then switch to your editor, then to a terminal, then to a chat with your assistant. Safeguard is built to meet you in each of those places with the same account behind it.
This post is a map. It lists each surface, what it is for, and how mature it is, so you can pick the one that fits the moment.
What it is
Safeguard is a platform for software supply chain security. The same account and the same data show up wherever you use it. Here are the surfaces, with their status stated plainly.
| Surface | Where | Status |
|---|---|---|
| Web app | app.safeguard.sh in a browser | Available |
| CLI | macOS, Windows, Linux | Available |
| Desktop app | Windows, macOS, Linux | Early access, beta |
| Mobile app | iOS and Android | Early access, beta |
| Chrome extensions | Chrome Web Store | Available |
| IDE extensions | VS Code, Open VSX (Cursor), JetBrains | Available |
| MCP server | mcp.safeguard.sh | Available |
The web app is chat-first. You can drive the platform by asking for what you want, and voice commands are supported. Everything else on the list connects back to that same account.
How to use it
Start where you already are.
In a browser. Sign in at app.safeguard.sh and ask a question in plain language, such as which of your projects have critical findings.
In a terminal. The Safeguard CLI scans a directory, repository, container image or SBOM on your own machine. It runs on macOS, Windows and Linux, installs from cli.safeguard.sh, and writes results as SARIF where a CI system expects them.
In your editor. The VS Code extension, also published on Open VSX so it works in Cursor, and the JetBrains plugin show security insight as you code. JetBrains covers IntelliJ, PyCharm, WebStorm, GoLand, Rider and others.
In your browser toolbar. Two Chrome extensions: Safeguard, which opens the Griffin search side panel, and Gold Open Source, which does package lookups.
In your AI assistant. The MCP server at mcp.safeguard.sh works with Claude, ChatGPT, Cursor or an agent you build. Tool calls are unlimited on every plan.
On desktop and phone. Both apps are in early access. Treat them as beta software and expect rough edges.
Admins stay in control
A tenant admin can switch off any client type: desktop, mobile, CLI, runners, IDE, browser extension, MCP clients, ChatGPT and Claude. If your team has a rule such as "no browser extensions", that rule can be enforced in one place instead of by memo.
Try it yourself
Pick two surfaces you do not use yet. Sign in once on the web app, then connect the second one, for example the CLI or an IDE extension. Ask both the same question and compare the answers. If the second surface is blocked, ask your tenant admin whether that client type has been switched off.
Quick check
Which surfaces are in beta today?
The desktop app and the mobile app are both early access, so treat them as beta.
Who decides which client types your team can use?
Tenant admins. They can switch off any client type, including the CLI, IDE, browser extension and MCP clients.
