Skip to content

RDC 0.3.0

Choose a tag to compare

@github-actions github-actions released this 09 Nov 12:03

RDC for AI agents, from anywhere, and the first Windows build.

  • From anywhere. The RDC window shows a second link, "From anywhere": it opens the viewer from GitHub Pages, finds the computer through public MQTT brokers (the topic and the encryption key come from the code and a key in the link) and connects to it directly over WebRTC — a phone on mobile data, another network, nothing to install. No servers of ours and no relay: if both networks block a direct path, the page says so. --no-remote turns it off.
  • AI agents. rdc mcp is an MCP server: an agent connects to a computer by the link from its RDC window, sees screenshots and uses the mouse and keyboard there (connect, screenshot, click, move, drag, scroll, type, key, wait, devices, disconnect). Works over stdio and over Streamable HTTP (--http), with a tunnel for cloud agents (--tunnel). Plugins: Claude Code (claude plugin marketplace add ANODs/RDC), Claude Desktop (rdc-mcp.mcpb below), Codex (codex plugin marketplace add ANODs/RDC), and a skill that teaches the agent how to work through RDC.
  • The person at the computer sees "AI agent …" in the RDC window and in the browser viewer and can stop it with one click.
  • Text input: viewers and agents type Unicode text whatever the keyboard layout on the host.
  • Viewer: the on-screen keyboard shows Latin or Cyrillic labels (the EN/RU button; 🌐 switches both the computer and the labels); on a phone, the whole page is the touchpad, not just the picture.
  • Windows: RDC.exe — the host on a PC (JPEG for now).
  • The whole project is in English.

Installation: macOS — unzip, move RDC.app to Applications and allow Screen Recording and Accessibility. Windows — download RDC.exe and run it. Claude Desktop — open rdc-mcp.mcpb.