Give Your AI Agent Eyes and Hands on a Remote Desktop with rdp-cli

A developer released rdp-cli, a command-line tool that lets AI agents control remote RDP and VNC desktops by capturing screenshots, clicking at coordinates, and typing text, then verifying each action with a follow-up screenshot. The tool runs on Linux x86_64 without additional runtime or administrator access, accepts passwords via stdin, and includes a human-readable viewer that starts read-only and supports pausing agent input or taking control of the session.

AI agents are good at generating text and calling APIs. But many real-world tasks still require interacting with graphical applications: editing a video, checking a spreadsheet, using legacy Windows software, or operating an Android phone. rdp-cli provides a simple command-line workflow for letting an AI agent control RDP and VNC desktops—and verify every action through screenshots. On Linux x86 64, download the executable and make it runnable: chmod u+x ./rdp-cli-linux-x86 64 ./rdp-cli-linux-x86 64 --help No additional runtime or administrator access is required. For Windows and macOS, see the quick start guide https://rdp-cli.com/en/docs/quick-start/ . Make sure RDP is enabled on the target computer. Then connect with your username and password: IFS= read -r -s -p 'Password: ' rdp password printf '\n' printf '%s' "$rdp password" | \ ./rdp-cli-linux-x86 64 connect \ --host 192.0.2.20 \ --user USER \ --password-stdin unset rdp password The command returns a session identifier. Save it as SESSION for the next steps. The password is passed through standard input rather than being written directly into the command line. Before taking an action, capture the current screen: ./rdp-cli-linux-x86 64 screenshot \ --session SESSION \ --out ./before.png An AI agent can inspect before.png , locate a button or input field, and decide what to do next. This observe → act → observe pattern helps the agent verify the current state after every action. After locating an input field at coordinates X, Y : ./rdp-cli-linux-x86 64 click \ --session SESSION \ --desktop \ --x X \ --y Y \ --observation OBSERVATION ./rdp-cli-linux-x86 64 screenshot \ --session SESSION \ --out ./focused.png ./rdp-cli-linux-x86 64 type \ --session SESSION \ --text "Remote task completed." \ --observation NEW OBSERVATION ./rdp-cli-linux-x86 64 screenshot \ --session SESSION \ --out ./after.png The agent can inspect the final screenshot instead of assuming the command succeeded. A person can open the same session in a native window or browser: ./rdp-cli-linux-x86 64 watch \ --session SESSION \ --viewer window The viewer starts in read-only mode and supports pausing agent input, taking control, and handing the session back. With rdp-cli, an AI agent can work with existing desktop applications and remote environments without requiring a new API integration for every tool. The basic loop is simple: Connect → Screenshot → Act → Screenshot → Verify Try the complete setup at rdp-cli.com https://rdp-cli.com/ . Always use authorized computers and trusted networks. Review the connection and certificate settings before using rdp-cli in production.