Remote Desktop Commander relays tool calls from AI applications to computers our users own. We take the security of that path seriously and appreciate responsible disclosure.

Remote Desktop Commander gives an AI assistant real file and terminal access on the machines you pair. Understanding the trust model matters:

• Tools execute with your user account's permissions on the paired machine. The service is an amplifier of whatever the connected AI client asks it to do — it can do anything your account can do.
• The connected AI account is trusted. Anyone who can drive your connected AI assistant (or your Remote Desktop Commander account) can reach your paired machines. Protect both with strong credentials and multi-factor authentication.
• Built-in restrictions are guardrails, not a sandbox. Configuration options like allowed directories and command blocking reduce accidental or unintended actions; they are not a security boundary against a malicious or compromised client.
• You are in control of exposure. A machine is reachable only while its device agent is running — stop it with Ctrl+C and access ends. Devices can be revoked at any time (individually or all at once) from the dashboard.
• Pairing is verified. Device pairing uses the OAuth device authorization flow: only approve a pairing code in the browser if it matches the code shown in your own terminal.
• Transport is encrypted in transit (HTTPS/TLS) and all relayed calls are scoped to your account.
• Relayed data is transient. Tool calls and their results pass through the hosted relay to deliver responses and are not retained once delivered; the service collects operational telemetry (event metadata such as tool name and success/failure — not file contents) as described in the Remote Desktop Commander privacy policy.

Recommendations: pair only machines you own; keep the agent running only when you need it; use MFA on your AI assistant and Remote Desktop Commander accounts; for high-risk or experimental work, run the agent inside a container or VM.

Please do not report security vulnerabilities through public GitHub issues.

Instead, use one of:

• GitHub private vulnerability reporting — Report a vulnerability on this repository (preferred)
• Email — legal@desktopcommander.app

Please include a description of the issue, steps to reproduce, and the potential impact. We will acknowledge your report within 3 business days.

• The hosted service at mcp.desktopcommander.app / auth.desktopcommander.app (OAuth flows, MCP endpoint, device pairing, session handling)
• The device agent (npx @wonderwhy-er/desktop-commander remote) — also reportable via the DesktopCommanderMCP repository, whose security policy covers the underlying open-source tool engine

• Accessing or modifying data belonging to other users
• Degrading the service for others (e.g. volumetric testing against production)
• Social engineering of Desktop Commander team members