Without WebMCP
Operate the ordinary page controls. The run document registers no WebMCP tools.
Start without WebMCPCONTROLLED DEMO / LOCAL MEASUREMENTS
One task. Two ways to complete it.
Compare the same booking application with WebMCP tools switched on or off.
SAME WEBSITE. SAME TASK.
Compare with and without WebMCP.
Use the same agent and model in a fresh conversation for each run. Both modes use the same calendar and confirmation flow.
Find the earliest available portfolio-review appointment on October 14, 2026, before noon UTC. Prepare it for my review and stop before confirming. Wait for me to confirm using the page. After I confirm, look up the reservation and report its reference, date, time and status.
With WebMCP
A supported browser agent can call the page’s structured tools. Ordinary controls stay available.
Start with WebMCPPreparing comparison…
The stopwatch starts when you press Begin timed run. It includes time switching to your agent. Time while a review is open is reported separately; it is not a measurement of the agent’s thinking time.
Compare results · 0 saved attempts
Latest attempt for each mode. These are local observations, not a verified AI benchmark. We cannot distinguish your clicks from an agent’s clicks.
| Observed measure | Without WebMCP | With WebMCP |
|---|---|---|
| Task outcome | — | — |
| Page stopwatch | — | — |
| Time with review open | — | — |
| Page time excluding review | — | — |
| Page-control actions | — | — |
| Native tool calls | — | — |
| Failed handler calls | — | — |
| Page confirmations | — | — |
| Execution evidence | No attempt recorded | No attempt recorded |
| Agent tokens, cost & total browser actions | Not available | Not available |
Up to 20 attempts are saved in this browser. DOM reads, all browser clicks, agent tokens and provider costs are not observed. No speed winner is inferred.
Checking browser support…Browser setup →