Connect an LLM Client to Your Site

Once your tools are live on Aigentably, AI agents that support WebMCP can call them directly. No API key sharing, no custom plugin required. This guide covers how to test and use your tools with Chrome's native WebMCP support.

How it works

Aigentably registers your tools via navigator.modelContext in the browser. When an AI agent navigates to your page, it discovers these tools and can call them on behalf of the user, without scraping or guessing.

WebMCP is currently in early preview in Chrome 146+ behind a flag. All other browsers are covered automatically by the Aigentably polyfill, so your tools still work. Agents just need a way to access them.

Enable WebMCP in Chrome

1. Open Chrome and navigate to chrome://flags/#enable-webmcp-testing
2. Set "WebMCP for testing" to Enabled
3. Relaunch Chrome

Install the Model Context Tool Inspector

Google provides an official Chrome extension for inspecting and calling WebMCP tools:

1. Search for "Model Context Tool Inspector" in the Chrome Web Store, or follow the link from the Chrome WebMCP early preview docs
2. Install the extension
3. Navigate to any page with Aigentably tools registered

The extension shows a badge when tools are available. Click it to see all registered tools with their name, description, and input schema. You can also call tools manually by filling in parameters and executing without an AI agent, or test with natural language by entering a prompt and letting the built-in Gemini agent pick the right tool.

Test your tools manually

The fastest way to verify your setup:

1. Navigate to your site (or http://localhost:3000/demo.html for the demo)
2. Open the extension
3. Select a tool from the dropdown
4. Fill in the input arguments as JSON, e.g.:

   { "productId": "shoe-01", "quantity": 1 }

5. Click Execute Tool. The result appears immediately.

This lets you confirm your executeJs is working correctly before involving any AI agent.

Test with natural language

The extension has a built-in Gemini agent for simulating real interactions:

1. Add your Gemini API key in the extension settings (get one free from Google AI Studio)
2. Type a natural language prompt, e.g. "Search for running shoes"
3. The agent picks the right tool, generates the correct arguments, and calls it

This is the best way to verify your tool descriptions and schemas are clear enough for an agent to use correctly.

Native agent support

Native WebMCP support in AI agents beyond Chrome's built-in testing extension is still emerging. As the standard matures, agents like Claude, ChatGPT, and Cursor are expected to adopt it. Your tools registered through Aigentably will work automatically when new agents add support.

Follow the WebMCP spec on GitHub to track adoption progress.