Reqable's MCP Server Lets AI Agents Run Packet Capture and Traffic Rewriting Hands-Free
How to Play with Packet Capture Testing in the AI Era
Author: MegatronKing Tags: Frontend, Backend, Testing
As the top domestic packet capture tool, Reqable has supported MCP calls in version 3.2, finally achieving integration with AI. Reqable + AI Agent, whether for API debugging or web testing, brings higher efficiency and more fun. Today I'm writing an article to briefly share how to play with packet capture testing in the AI era.
1. Tool Preparation
First, any AI Agent — whether Codex, Claude Code, Cursor, or Copilot, Windsurf, or TRAE — that supports configuring an MCP server will work. Here we choose the built-in Github Copilot in VS Code for demonstration, with the model GPT-5.4. The choice of Agent and model doesn't matter much; don't overthink it. Since only a small amount of code writing is involved, basically any will suffice.
Then, download and install the latest version 3.2.7 from the Reqable official website. After initialization, complete the certificate installation with one click, then configure Reqable's MCP server in VS Code. Reqable already provides MCP server configuration guides for various AI Agents. Open the app menu Tools -> MCP -> VS Code, and click Auto Configure to automatically open VS Code and install.
Click Install in VS Code, and the MCP server is configured. After installation, use Ctrl + Shift + P and type the MCP keyword to find MCP:List Servers and check if the server is already Running.
Reqable MCP (open-sourced on Github, address at the end of the article) provides over a hundred Tools for AI to control, but by default only commonly used Tools are registered. When configuring MCP, you can enable all by using the parameter --scope all.
Note that having too many Tools can bloat the context. Even if all Tools are registered, many AI Agents won't load them all; you need to actively tell them to load everything. We'll stick with the default configuration here and only load the necessary Tools.
To let AI control Chrome, we also additionally installed Chrome Devtools MCP. The entire mcp.json configuration file is as follows:
{
"servers": {
"chrome-devtools": {
"command": "npx",
"args": [
"-y",
"chrome-devtools-mcp@latest"
],
"type": "stdio"
},
"reqable": {
"type": "stdio",
"command": "/Applications/Reqable.app/Contents/Helpers/mcp-server",
"args": []
}
}
}
Next, we hand both Reqable and Chrome over to AI control.
2. Automatic Packet Capture
First, we give AI a simple test instruction, as follows:
Start Reqable packet capture, then use Chrome to open the reqable.com webpage, making sure to disable browser cache.
AI will automatically control Reqable to configure the system proxy and start packet capture, and control Chrome to automatically open the reqable.com webpage in cacheless mode. After the page loads, it will automatically check whether Reqable captured the web requests. We can also switch to the Reqable window to confirm.
We continue giving AI instructions, letting it specify which request in Reqable to analyze. You can specify an ID, a URL, or even various filter conditions, such as requests containing the keyword Hello World.
Analyze the request with ID 386 in Reqable.
Besides analyzing requests, you can also let AI create API tests in Reqable, generate cURL, add to specified API collections, and more, making it convenient for developers and testers to conduct follow-up investigations.
Create an API test for this request in Reqable.
If necessary, you can perform API testing right in Reqable, completely eliminating the need to manually edit and import APIs. Postman and the like can all go collect dust.
3. Data Debugging
Below is a simple demonstration of some advanced features, such as rewriting and scripting. During development and testing, there are often needs for data modification, such as changing location coordinates or modifying order amounts. The common operation is to open a packet capture tool, write modification rules, and then test. In the AI era, these manual operations can all be eliminated.
We'll continue based on the session above, letting AI control things, and simply simulate such a scenario.
Create a rewrite rule to change all instances of the character string Reqable in the reqable.com website content to Awesome
As can be seen, AI automatically did the following things for me, fully automated throughout the process.
- Created a rewrite rule in Reqable, string replacement Reqable -> Awesome
- Enabled the rewrite feature switch and the newly created rule in Reqable.
- Refreshed the webpage and checked whether the modification took effect.
Switch to the browser to check, and the copy has indeed been changed.
Besides letting AI control rewriting, you can also let it control request breakpoints, write scripts to process requests, and more. Below is a demonstration of the automatic script writing feature, with the following instruction:
Write a Reqable script and enable it.
Save all image resources from the website reqable.com to the current user's Downloads directory, naming the save folder after the domain name.
Refresh the webpage to let the script execute.
Completely without manual operation, we see that the images have been successfully saved according to the Path file structure. Let's also open the script in Reqable to glance at the code quality.
Very simple Python code, but writing it by hand would still take considerable effort; AI handles it in one go. Additionally, we see two pleasant surprises at the end that go beyond the requirements!
context.comment = f'Saved image to {final_path}'
context.highlight = Highlight.green
To briefly explain, the first line adds a comment to the image request that triggered the save, and the second line sets a green highlight. AI is so thoughtful. Below we can check in Reqable.
From thousands of requests, we quickly found the green-highlighted request, and checking the comment reveals the full path where this request was saved 🐮🍺.
4. Conclusion
In the AI era, many purely manual tasks will gradually be replaced by AI, not just programming but also many testing tasks. As Confucius said, "A craftsman who wishes to do his work well must first sharpen his tools." Good tools get twice the result with half the effort. Reqable, as a well-known domestic packet capture tool, is worth having.
Reqable MCP project open-source address: