Skip to main content

Webcmd

Webcmd gives coding agents a real browser for public and authenticated web tasks. As an agent explores a site, Webcmd retains useful structural knowledge locally so later tasks can begin with context instead of rediscovering the same interface.

Add Webcmd to Your Agent

For Codex, open Plugins, choose Add plugin marketplace, enter agentrhq/webcmd, and install Webcmd. Start a new task after installation. On first use, the integration installs the npm CLI automatically if webcmd is missing. For Claude, another supported harness, or a standalone Codex setup:
Choose one Codex skills-delivery path so the webcmd-browser skill is not installed twice.

Then Ask for an Outcome

Describe the result you need, the fields to return, the browser profile to use, and anything the agent must not do.

What Your Agent Can Do

Your agent can navigate unfamiliar websites, work inside an authenticated browser profile, extract structured results, pause for human sign-in, and reuse verified knowledge of a site’s layout on later runs. The live page remains the source of truth when a site changes.

Where to Go Next

Continue with the Quickstart, then copy a task from the Prompt Cookbook. Read How Webcmd Works for the browser, Session, and self-learning model.