mcpX product documentation
Run your first workflow
Use clear browser tasks, explicit tabs, and fresh page snapshots with mcpX.
Once the server and extension show as connected, ask your MCP client to use mcpX. Good requests name the target, the desired action, and the evidence you want back.
Start with a read-only task
Open a page, then ask:
Use mcpX on my active tab. Read the page structure and list its main headings and links. Do not click anything.This usually calls tab-current followed by snapshot and is a safe way to
verify the connection.
Run an interaction
Use mcpX to open example.com in a new tab, inspect the page, and follow the More information link. Tell me the final URL.The client can combine open-tab, snapshot, tab-click, and tab-current.
Ask for evidence
For testing or debugging, say what should prove the task succeeded:
Fill the signup form with test data, submit it, wait for the confirmation message, and capture a screenshot of the result.Reliable task patterns
Name the tab
When multiple tabs are open, include the site or ask the client to list tabs first. Explicit tab targeting reduces accidental actions in the wrong page.
Inspect before interacting
Use snapshot to obtain semantic elements and generation-scoped references.
After navigation or a large page update, take a new snapshot before the next
click or input. Old references intentionally expire.
Wait for observable state
Prefer a condition such as “wait for the success message” or “wait for
navigation” over a fixed delay. The wait tool supports bounded navigation and
element-state waits.
Treat page actions as real actions
mcpX can click destructive controls, submit forms, and type into signed-in pages. Review consequential steps and ask the client to pause before purchases, messages, deletions, or account changes.
See Tool reference for the complete browser surface.