With Custom MCP Tools, developers can expose structured information and actions directly to agents, replacing repeated page analysis with purpose-built tools. The result is dramatically faster workflows, significantly lower token consumption, and more reliable automation.
In the example below, an AI agent completes a multi-step assessment journey on an existing Dutch VvE (homeowners association) sustainability helpdesk website. The user simply describes their situation and goal. The agent completes the assessment, reviews the recommendations, and emails the report.
This runs on an existing website and workflow. No code changes were required.
|
|
Without Custom MCP Tools |
With Custom MCP Tools |
|---|---|---|
|
Token usage |
~66,000 |
~3,000 |
|
Completion time |
~3 minutes |
Under 1 minute |
That's a 95% reduction in token consumption and a workflow that completes more than three times faster.
Want to build your own tools? Read our guide to Custom MCP Tools:
https://dev.webfu.se/custom-mcp-tools/
The same approach can make AI-powered customer journeys significantly faster and more cost-efficient by reducing the amount of context agents need to process. Think of workflows like:
-
Buying or modifying an insurance policy
-
Opening a bank account and completing KYC
-
Filing taxes across government and third-party portals
All on top of existing applications and websites. No code changes required. Third-party systems can be part of the same workflow, with every action remaining fully auditable.
Sometimes the biggest improvement isn't a better model. It's giving the model the right tools.