TestChimp Local Agent - Agentic Testing, from your CLI

Introduction

TestChimp is an AI QA platform, that does autonomous agentic explorations of webapps to:

• Build a comprehensive mindmap about the webapps' behaviour (Capturing different screens, states, pathways, down to individual elements in screens)
• Captures critical bugs across visual, security, functional, UX, usability, accessibility and many more categories by analyzing DOM, Screenshots, Network and Console logs.
• Auto-generate test scenarios for the webapp
• Write test scripts for the identified test scenarios
• Provide visual feedback for screens in your webapp

The mindmap built by TestChimp is then made available as contextual knowledge for various QA workflows such as:

• Scoping out detailed user stories
• Brainstorming test scenarios to consider
• Writing precise and detailed test steps for test scenarios
• and more...

You can sign up for free at: https://testchimp.io

TestChimp-Local

When you trigger an exploration from testchimp site, the agent runs on TestChimps cloud. testchimp-local enables you to run TestChimp's autonomous testing agent locally. This enables the following:

• Test your localhost version (during development phase, before even pushing to staging envs)
• Reuse your signed-in sessions for testing sites with complex logins (eg: SSO, 2FA etc.)
• When your site under test is not open to public internet access (behind firewalls etc.)
• For human-in-the-loop hybrid exploration with the agent.

This also doubles-up as an MCP server that MCP clients (such as Cursor, Windsurf, Claude Desktop etc.) can integrate with for tool usage. Refer to MCP server setup section below for more details.

Prerequisites

Python 3.11 or above should be installed.

How to Run

1. Install:

pip install --upgrade testchimp-local

2. Create an exploration configuration file

Refer Configuration section below for guide and examples.

Here is a minimal config file to get started:

{
  "browserContext": {
    "mode": "launch",
  },
  "appReleaseLabel": "local_default",
  "explorationConfig": {
    "promptConfig": {
      "url": "https://example.com",
      "explorePrompt": "Explore the website and find any issues",
      "testDataPrompt": "Login with test@example.com password1"
    },
    "maxCredits": 50,
    "maxJourneys": 2,
    "bugCaptureSettings": {
      "sources": ["dom", "screenshot", "network", "console"]
    },
    "viewportConfig": {
      "viewports": [
        {
          "nickname": "laptop"
        }
      ]
    },
    "urlRegexToCapture": ".*\\.example\\.com.*"
  }
}

3. Set environment variables

Following environment variables are used for authenticating your local agent client and syncing with TestChimp platform:

• TESTCHIMP_PAT: Sign in to your TestChimp account. Go to user settings (click on your name at side bar bottom). Copy Personal Access Token value.
• TESTCHIMP_PROJECT_ID: The project id for which the results will be synced. Go to project settings -> General. Copy project id.
• TESTCHIMP_EMAIL: Email address used for signing-up with TestChimp

3. Run

testchimp-local --config_file=<path to your exploration_config.json>

By default, the local service is started at port 43449. You can change it be passing --port argument.

Configuration

Here's a comprehensive guide to all available configuration fields:

Top-Level Configuration

{
  // Optional (default: launch mode). Browser context to use. You can connect to an existing browser instance via cdp, or let agent start a new instance (launch).
  "browserContext": { /* browser configuration */ },
  // Required. The configuration file detailing settings for the exploration
  "explorationConfig": { /* exploration configuration */ },
  // Optional. The release label for the site being tested. Bugs identified, test scenarios generated, will be tagged with this release in TestChimp platform.
  "appReleaseLabel": "local_default"
}

Browser Context Configuration

{
  "browserContext": {
    "mode": "launch" | "cdp",
    "cdpUrl": "http://localhost:9222",
    "headless": true | false
  }
}

• mode: Browser launch mode
  • "launch" - Launch a new browser instance
  • "cdp" - Connect to existing browser via Chrome DevTools Protocol
• cdpUrl: Chrome DevTools Protocol URL (required when mode is "cdp")
• headless: Whether to run browser in headless mode

Exploration Configuration

{
  "explorationConfig": {
    "promptConfig": { /* prompt-based configuration */ },
    "scriptConfig":{ /* script-based configuration */},
    "maxCredits": 100,
    "maxJourneys": 5,
    "bugCaptureSettings": { /* bug capture configuration */ },
    "viewportConfig": { /* viewport configuration */ },
    "urlRegexToCapture": ".*\\.example\\.com.*"
  }
}

Prompt-Based Configuration

This is for when you want the agent to autonomously explore the webapp. You can provide (optionally) an exploration focus (via the explorePrompt), and the test inputs to use (via the testDataPrompt).

{
  "promptConfig": {
    // Required: URL to visit
    "url": "https://example.com",
    // Optional - any additional instructions for context
    "explorePrompt": "Explore the website and find any issues",
    // Optional - any specific test input values to use
    "testDataPrompt": "Login: test@example.com pass: Password1"
  }
}

• url: Required. Target website URL to explore (The starting point of the exploration)
• explorePrompt: Optional. Instructions for the AI agent on what to explore
• testDataPrompt: Optional. Instructions for any specific test input values to use

Script-Based Configuration

A.ka. the "Guided Mode" - this is for when you want the agent to follow specific journeys, by providing playwright js / ts scripts as guiding steps. The agent will still analyze the screenshots, DOM, network and console for bugs, write test scenarios, generate scripts and build the mindmap while following the guided steps through the app.

{
  "scriptConfig": {
    // Required: list of file paths for scripts to run
    "filePaths": ["src/tests/home_visit.spec.js","src/tests/successful_login.spec.js"]
  }
}

Bug Capture Settings

{
  "bugCaptureSettings": {
    "sources": ["dom", "screenshot", "network", "console"],
  }
}

• sources: Data sources to analyze (string values supported):
  • "dom" - DOM analysis for visual issues
  • "screenshot" - Screenshot analysis for visual issues
  • "network" - Network request/response analysis
  • "console" - Console log analysis

Viewport Configuration

{
  "viewportConfig": {
    "viewports": [
      {
        "nickname": "laptop" | "widescreen" | "mobile" | "tablet",
        "width": 1366,
        "height": 768
      }
    ]
  }
}

• nickname: Predefined viewport size (string values supported):
  • "laptop" - Standard laptop resolution
  • "widescreen" - Wide screen resolution
  • "mobile" - Mobile device resolution
  • "tablet" - Tablet device resolution
• width: Optional. Custom viewport width in pixels
• height: Optional. Custom viewport height in pixels

Complete Example Configuration

{
  "browserContext": {
    "mode": "cdp",
    "cdp_url": "http://localhost:9222",
    "headless": false
  },
  "explorationConfig": {
    "promptConfig": {
      "url": "https://example.com",
      "explorePrompt": "Explore the website and find any issues",
      "testDataPrompt": "Login with test@example.com password1"
    },
    "maxCredits": 100,
    "maxJourneys": 5,
    "bugCaptureSettings": {
      "sources": ["dom", "screenshot", "network", "console"]
    },
    "viewportConfig": {
      "viewports": [
        {
          "nickname": "laptop"
        }
      ]
    },
    "urlRegexToCapture": ".*\\.example\\.com.*"
  },
  "appReleaseLabel": "local_default"
}

Reusing an existing browser session

To reuse your existing signed-in sessions (to circumvent 2FA, SSO limitation etc.) for agentic testing, follow the steps below:

1. Close your existing chrome browser instances.
2. Start chrome in CDP mode (below command):

In MacOS:

/Applications/Google\ Chrome.app/Contents/MacOS/Google\ Chrome --remote-debugging-port=9222 --user-data-dir=/tmp/chrome-cdp

In Windows:

"C:\Program Files\Google\Chrome\Application\chrome.exe" --remote-debugging-port=9222 --user-data-dir=C:\chrome-cdp

In Linux:

google-chrome --remote-debugging-port=9222 --user-data-dir=/tmp/chrome-cdp

3. In your exploration config json file, specify the browser_context section as below, and it will connect to the browser instance via CDP, and reuse the existing sessions:

  "browserContext": {
    "mode": "cdp",
    "cdpUrl": "http://localhost:9222",
    "headless": false
  }

Using as an MCP server

testchimp-local as an MCP server as well, enabling the following capabilities:

1. Provide access to your webapps' runtime behaviour context - capturing the different screens, pathways, states and detailed down to individual element level granularity, giving your IDEs deep contextual awareness about the product being built.

2. Visual Vibe Coding: Point and just "Say the change you want to see" in the browser.

3. Fetch recent console logs from browser (for context-aware troubleshooting, checking fixes implemented are working etc.)

4. Run agentic exploratory testing

Prerequisites

1. Install TestChimp Chrome extension

TestChimp MCP server communicates with the TestChimp chrome extension to fetch current browser state information for quick feedback visual vibe coding sessions.

2. Install TestChimp VSCode extension

The VSCode extension acts as a communication bridge between VSCode and the browser extension.

Installation and usage

1. Install the server as instructed above.

pip install --upgrade testchimp-local

2. Set environment variables

Following environment variables are used for authenticating your local agent client and syncing with TestChimp platform:

• TESTCHIMP_PAT: Sign in to your TestChimp account. Go to user settings (click on your name at side bar bottom). Copy Personal Access Token value.
• TESTCHIMP_PROJECT_ID: The project id for which the results will be synced. Go to project settings -> General. Copy project id.
• TESTCHIMP_EMAIL: Email address used for signing-up with TestChimp

3. Run the server

testchimp-local

It starts in port: 43449 by default. You can change to a different one by passing --port argument.

4. Configure Cursor / Windsurf to use this as an MCP server by adding following to mcpServers config file:

    "testchimp":{
      "url":"http://localhost:43449/mcp/"
    }

5. Your MCP client (Cursor, Claude Desktop etc.) can now use the TestChimp MCP server to leverage the above detailed capabilities. To see it in action, open your webapp in the browser, and open the TestChimp chrome extension for visual vibe coding, capturing bugs, resolving bugs right from the browser and more...