CARBON by Testers.ai

What it does

CARBON by Testers.ai is an AI testing harness used inside a supported coding-agent environment. Developers and QA practitioners can point it at a project or change and ask it to assess risk, explore user flows, investigate accessibility, or test a named feature. It works through the tools and model access configured in the host agent; it does not replace an existing test framework.

The harness analyzes project context and test evidence, suggests tests to run or create, and produces findings with reproduction steps, reports and coverage gaps. Free Core and paid Pro editions differ in scope; confirm current package entitlements before relying on a particular workflow. Testing is bounded by the selected scope, host permissions and budget. Generated tests need execution, and a property proof does not certify an entire application.

Install a native plugin for Claude or Codex, or connect through a local stdio MCP server in a supported coding-agent host. Node.js 22 or newer is required; demo and full-workflow helpers also need Python 3.10 or newer. The host agent may send authorized code and evidence to its model provider. Model access and usage charges are separate from CARBON licensing.

What it can do

• Risk-based project and change assessment
• Exploratory testing of user flows
• Accessibility investigations
• Reproducible bug findings and regression-test proposals
• Evidence-backed reports with coverage gaps
• Specialist investigations and Lean-backed property proofs in Pro

Works with

• Claude
• Codex
• Cursor
• Antigravity IDE
• MCP
• Playwright
• Selenium
• pytest
• Appium
• Jira

Before you choose

• Your coding agent supplies model access; subscription allowances or API charges are separate from CARBON licensing.
• The coding agent may send authorized code, screenshots and prompts to its model provider; local reports do not mean offline or local-only AI.
• Scope, edition, host permissions and budgets limit testing; generated tests need appropriate dependencies and execution, and Lean proofs do not establish whole-application correctness.

Listed under

• Coding
• Accessibility testing
• AI testing
• AI test manager
• Coding agents
• Developers
• QA

Get started

The official install page lists Claude, Codex, Cursor and other MCP hosts.

1. Choose one integration: Claude or Codex native plugin, or MCP for Cursor and other supported hosts.
2. Install the plugin or configure the local stdio MCP server. Use Node.js 22+; demo and full-workflow helpers also require Python 3.10+.
3. Open a project in the coding agent and give Jay a specific test request; inspect the evidence and coverage gaps in the resulting report.

Live official checkout lists Pro at USD49/month and Team at USD95/developer/month, minimum five developers (USD475/month). Model usage is separate. Confirm current package capabilities and delivery before purchase; documentation and homepage differ on edition scope.

Source: testers.ai, checked 2 Nov 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 2 Nov 2026. Each source supports part of the overview; availability and prices can change.

• CARBON, powered by testers.ai — AI testing platform ↗testers.ai
• Install CARBON for Claude, Codex, Cursor or MCP | CARBON by testers.ai ↗testers.ai
• CARBON, powered by testers.ai — AI testing platform ↗testers.ai
• CARBON Free documentation · CARBON ↗testers.ai

Company website

testers.ai: product, pricing and policy pages

YC directory mirror

Batch and company name

Human review

Checked by a person

knowledge_base

manual_review

Last checked 2 Nov 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.