CARBON AGENT · CORE

Meet Jay.
Your AI test manager agent.

An AI test harness that lives in your coding agent. Jay coordinates 12 free testing commands—from risk-based checks and exploratory testing to accessibility investigations, reproducible findings and evidence-backed confidence.

Works alongside your existing test frameworks and your coding agent’s tools. Free and open source.

Install Core free →Explore the 12 commands ↓

JAY + YOUR CODING AGENT

Your testing, coordinated.

12commands ↓

MITfree & open source

Risk-based testing Exploratory checks Accessibility Evidence & reports

Supported

CARBON can use any connection established in your AI coding agent context.

Clients & protocol

Claude

ChatGPT

Codex

Cursor

MCP

Frameworks & tools

Selenium

Playwright

pytest

AppiumVibium

Languages

Java

Python

TypeScript

JavaScript

Issues

Jira

GitHub Issues

GitLab Issues

Test management

TestRailXray

Browsers & platforms

Chrome

Firefox

Safari

iOS

Android

Device & cloud testingBrowserStack

Sauce Labs

ONE PLUGIN. YOUR EXISTING WORKFLOW.

Choose your agent.

Install CARBON. Open your project. Give Jay one thing to test.

Step 1Add to your agent

Open /plugin and choose Discover.
Search CARBON. Check that the publisher is testers.ai, then install.
Start a fresh session and choose /carbon:carbon.

In Claude Cowork, use the plugin directory. Your organization may restrict plugins.

Installation help →

Step 1Add to your agent

Open the listing, verify the testers.ai publisher, and install in your supported local Codex environment. Start a fresh session and choose CARBON in the skill picker.

Open CARBON in the marketplace ↗

The local plugin needs your coding agent’s tools. Ordinary ChatGPT web chat is not the local Codex runtime.

Step 1Add to your agent

Add CARBON as a local MCP server in Cursor using this configuration, then reconnect and check that its tools are available.

{
  "mcpServers": {
    "carbon": {
      "command": "npx",
      "args": [
        "-y",
        "@testers-ai/carbon-mcp"
      ]
    }
  }
}

Requires Node.js 22+. This runs a local stdio server, not a hosted URL. Some Windows hosts need npx.cmd. Merge the CARBON entry into your existing config; don’t replace other servers.

Once connected, ask your agent to use CARBON to test your project. MCP does not automatically add the native plugin’s slash commands.

Package & setup details ↗

Step 1Add to your agent

In Antigravity IDE, open … → MCP Servers → Manage MCP Servers → View raw config.
Merge the configuration below into mcp_config.json and save.
Refresh the MCP servers and check that CARBON connects before asking it to test.

{
  "mcpServers": {
    "carbon": {
      "command": "npx",
      "args": [
        "-y",
        "@testers-ai/carbon-mcp"
      ]
    }
  }
}

Requires Node.js 22+. This runs a local stdio server, not a hosted URL. Some Windows hosts need npx.cmd. Merge the CARBON entry into your existing config; don’t replace other servers.

Once connected, ask your agent to use CARBON to test your project. MCP does not automatically add the native plugin’s slash commands.

Official Antigravity MCP setup ↗

Step 1Add to your agent

Use any MCP host that supports local stdio servers. Server name: carbon; command: npx; arguments: -y @testers-ai/carbon-mcp.

Paste the JSON below if your host uses an mcpServers configuration. Otherwise enter the command and arguments in its MCP settings. Save, reconnect and check the available tools.

{
  "mcpServers": {
    "carbon": {
      "command": "npx",
      "args": [
        "-y",
        "@testers-ai/carbon-mcp"
      ]
    }
  }
}

Requires Node.js 22+. This runs a local stdio server, not a hosted URL. Some Windows hosts need npx.cmd. Merge the CARBON entry into your existing config; don’t replace other servers.

Once connected, ask your agent to use CARBON to test your project. MCP does not automatically add the native plugin’s slash commands.

CARBON MCP package ↗

More than “the tests passed.”

Test the change. Explore the experience. See what remains uncertain—without replacing your existing framework.

Test what matters.

Prioritize high-risk changes and explore real user flows. Look beyond happy paths to find where the experience breaks.

Risk analysis · Exploratory testing

Find it. Reproduce it.

Keep the steps and evidence behind verified bugs. Get regression-test proposals to help stop the same failure returning.

Reproducible findings · Regression tests

See the evidence.

Inspect readable reports with findings, supporting evidence and coverage gaps. Know what was checked—and what wasn’t.

Rendered reports · Coverage gaps

Go further with `/full`.

Run the comprehensive testing workflow across your project. Included free, within your edition’s capabilities and your agent’s permissions.

Comprehensive workflow · Included free

INCLUDED FREE · MIT OPEN SOURCE

Your 12-command toolkit.

Click a command to watch its walkthrough. Use Copy command or copy an example below. Replace sample paths and features with your own.

Run a bounded risk-based assessment of a project or URL, including business value, stateful checks, persona exploration and an evidence-qualified report.