WebMCP - Model Context Tool Inspector

A Chrome extension for inspecting, monitoring, and executing tools exposed via the experimental navigator.modelContextTesting Web API, with optional AI-assisted interaction powered by a local LLM via LM Studio.

Prerequisites

• Chrome 146.0.7672.0 or higher — enable the WebMCP for testing flag at chrome://flags/#enable-web-mcp.
• LM Studio (optional) — required for the AI prompt/chat features. A model that supports OpenAI-style tool calling is recommended.

Installation

This extension is not on the Chrome Web Store. Install it locally from source.

1. Clone the repo:

   git clone <repo-url>
   cd model-context-tool-inspector

2. Build the bundle:

   npm install

   This bundles the OpenAI SDK into js-openai.js and removes node_modules.

3. Load into Chrome:

   • Navigate to chrome://extensions/
   • Toggle Developer mode (top right)
   • Click Load unpacked and select this directory

Usage

Inspecting tools

1. Open a page that exposes WebMCP tools.
2. Click the extension icon in the toolbar to open the Side Panel.
3. The panel lists all tools the page exposes — double-click a row to pretty-print it.
4. Select a tool, edit the input arguments, and click Execute Tool to call it directly.

AI-assisted interaction

The panel can send natural-language prompts to a local LLM that automatically calls WebMCP tools on your behalf.

1. Start LM Studio and load a model.

2. Start the local server — in the LM Studio UI go to the Developer tab and click Start Server, or run:

   lms server start

   The server defaults to http://localhost:1234.

3. Open the advanced menu (⋮ button next to the prompt) in the sidebar:

   • Confirm the endpoint is http://localhost:1234/v1 (change it if your LM Studio runs on a different port).
   • The model dropdown will populate automatically. Select the model you want.

4. Type a prompt and press Send (or Enter). The LLM will call tools as needed and return a result.

5. Reset clears the conversation history so the next prompt starts fresh.

6. Copy trace copies the full request/response trace to the clipboard for debugging.

Rebuilding

If you update the openai SDK version in package.json, re-run npm install to regenerate js-openai.js.

Disclaimer

This is not an officially supported Google product.