Not everyone wants a terminal. Safeguard Code, the local coding agent that writes and refactors with security in mind, is available on desktop as well as in the command line, so you can work with it in a window beside the Safeguard web app.
What it is
The same agent that runs in the terminal also works from the Safeguard desktop app, which runs on Windows, macOS and Linux. The desktop window has the web app on the left and a local panel on the right. The panel has tabs for Scan, Terminal, MCP, Tasks, Files and Settings, and it can reach things a browser tab never can, such as the folders you choose to share.
That combination is the point. The agent sits next to a view of your findings, your scan results and your folders. You ask for a change, watch what it does, and check it against the project's policy without switching tools.
What you do with it
- Ask for a fix in plain language. Fix a reachable CVE, upgrade a library across a project, or review a change before commit.
- Scan without uploading. Add a local project and scan it. Results appear in the panel, grouped from critical down.
- Run tasks with approval. The Tasks tab runs multi step work in sequence and pauses for your approval before anything destructive.
- Work with MCP tools. Start a Safeguard MCP server from the panel, or connect another, and see which tools an agent can call.
- Keep control of files. The app only reads folders you grant, and the terminal only runs a fixed list of programs.
Why it matters
A coding agent is only as safe as the boundaries around it. In the desktop app those boundaries are visible: which folders are granted, which programs may run, what is waiting for approval. For a team that mixes terminal users with people who prefer a window, one agent in both places means the same behavior and the same trace.
Example
A developer adds a service folder, scans it, and sees a reachable vulnerability in an indirect dependency. They ask the agent to upgrade it. The agent proposes the change, runs the build and tests, and checks the result against policy. The developer reads the diff in the panel and commits.
How to start
- Download the desktop app from the downloads page and sign in.
- Grant one project folder.
- Run a scan, then ask the agent for one small change.
- Review the result as you would any pull request.
Quick check
Why run the agent beside the web app?
You can see findings, scan results and granted folders while the agent works, so you can check each change against policy.
