domains / families / agent-tooling / webmcp
webmcp
separate
Open a user-provided URL in the host's built-in browser and use the page's MCP or WebMCP tools before browser UI automation for app communication or edits.
When to use
- Open a web app in the host's built-in browser
- Use page's MCP or WebMCP tools
- Complete requests through page tools
- Navigate and inspect visually
- Handle sign-in prompts
- Execute tool calls via document.modelContext
When not to use
- External Chrome window
- Normal tab
- Browser extension
- Click, type, drag, or screenshot for tool-capable operations
- DOM automation for tool-capable operations
- Enter, copy, inspect, or request passwords, cookies, tokens, or verification codes
- Substitute generic tool-search or remote API for current tab's page tools
- Copy tool descriptors out of the page to invoke from host
Skill page
Source