Testers.ai CARBON Test Harness Lite

CARBON Core — free, open-source AI testing for Codex

Your AI test manager agent. Jay prioritizes risk, investigates bugs, checks accessibility and helps create regression tests using your existing framework. Run `/carbon` and follow the visual execution plan, current check and recorded screenshots. Inspect findings, evidence, confidence qualifications and coverage gaps in the live local report. Free under the MIT license.

Website and guide · Download releases · Claude edition · Report an issue

Your agent's plan. Your local workspace.

Use the model access you already have. In Claude Code, Codex, and other coding agents, CARBON's default testing workflows use the host's configured model access. When the host uses an eligible subscription, CARBON uses that allowance—no separate CARBON model API key is required. If the host uses API credentials or a company LLM proxy, that billing and data policy apply instead. Plan limits and extra-usage charges still apply; CARBON Pro licensing is separate.

Local execution and local reports—not a promise that no data leaves your machine. For local runs, project files, test evidence, and reports are stored in your local workspace. Your coding agent can send selected code, screenshots, prompts, and other context to its configured model provider. Local CARBON is not automatically offline or air-gapped. Cloud runs, external integrations, and optional direct-model checks have separate data flows and may have separate costs. Command-usage analytics vary by edition; review that edition's privacy policy and settings.

To confirm your billing route, check /status in Claude Code or codex login status in Codex. A configured API key can select API billing instead of subscription usage. Standalone MCP clients must supply their own model access and execution tools; installing an MCP server does not include a model subscription.

Software testing, accessibility testing, bug finding, regression test planning and quality reports with Jay, your AI testing agent. Open source under MIT, readable JavaScript. The Codex plugin needs no MCP server, OAuth, account, hosted testing service or CARBON model calls. Built by testers.ai. This is a separate edition; it does not replace the existing Claude package or the original Studio submission. The separate private CARBON Studio Pro edition preserves the full local MCP experience. Carbon Core updates the existing published plugin; its package identifier remains carbon-lite for compatibility.

Standalone MCP distribution

For MCP clients and directories, use the separate local MCP bundle. It adds stdio tools and prompts around the Carbon Core evidence engine, not Pro features. The Codex plugin ZIP installs no MCP server. Its explicit view action opens a temporary, protected, read-only loopback report viewer (30-minute lifetime); evidence actions remain serverless.

Install in Codex

Download the ZIP from Releases, extract it, and use your Codex version's local-plugin installation flow to select that folder. This repository is not a claim of public directory approval. After installation, start a fresh conversation and run `/carbon` (select CARBON's carbon command in the host's skill picker if it requires a plugin prefix). CARBON tests the current project when the target is clear. Otherwise it asks what to test: provide a URL, folder, repository, API, app, feature, or requirements. It runs a broad assessment within the configured budget and reports findings, evidence-qualified confidence, coverage gaps, and prioritized next steps. Host capabilities and permissions determine which tests can actually run.

License

CARBON is open source under the MIT license. You may use it at work, fork it, modify it, redistribute it, and sell copies, preserving the required copyright and license notices. Third-party dependency licenses remain intact; see NOTICE.md. There is no trademark endorsement grant.

Commands

Invocation prefixes depend on the host. A useful first request is: “Use CARBON to assess this project and show the evidence report.”

Skills can analyze supplied requirements or results without a local runtime. Hands-on testing needs the host's browser/terminal/API capabilities. The optional evidence helper requires Node 22+, already bundled as readable JavaScript: no npm install for archive users. Demo artifacts may require Python or their documented test framework; creating a copy neither installs dependencies nor starts a server.

What is and is not equivalent

Included: Core investigation methods, disposable demos with test profiles, local preferences, revisioned evidence/history, findings, persona journeys, screenshots, scope-qualified confidence, portable HTML/JSON, screenshot enlargement, map/check inspection and explicit feedback import/export.

Different from MCP Studio: no native global/settings entrypoint, MCP tools, automatic host message handoff, unattended UI write-back, or guaranteed in-chat live rendering. The portable report is an HTML artifact. The protected local viewer polls recorded state for live plans, checks and screenshots; it does not independently execute tests. UI feedback is a draft until explicitly imported. Not included: Pro commands, proprietary benchmarks, hosted sharing, cloud execution, Pro specialist commands or hosted sharing. The included /full workflow is resumable; bounded local quick checks are available only with trusted manual-install hooks.

Cost and privacy

No infrastructure bill for this edition. Host subscriptions/API use, local compute and any test-target costs remain separate. State stays in the selected project's .carbon/studio-lite/; no hidden account or global configuration is created. Outputs are private local files but contain any evidence recorded. Review before sharing. The host model provider's processing is separate from this plugin. No automatic retention/deletion: remove only the specific project evidence folder when you choose. Never remove other CARBON state as part of uninstalling this edition.

Develop and package

The display name is CARBON Core; carbon-lite remains the stable technical plugin identity so existing installations and the marketplace listing can receive updates. The listing emphasizes software testing, QA, risk-based tests, bug finding, accessibility and evidence-backed reports rather than edition size.

Use npm run package:marketplace for the public OpenAI directory. That produces a separate -marketplace.zip without lifecycle hooks. Users explicitly invoke its testing commands; it does not automatically check code after edits. The ordinary npm run package output retains hooks for supported manual installations. Do not upload the manual-install ZIP to the public directory.

npm ci

npm run package

The ZIP contains the manifest, skills, runnable helper, full source, bundled testing reference catalogs, fixture code and third-party notices. It excludes node_modules, local evidence, test outputs, credentials and private benchmark data.

See SUBMISSION.md for the local skills/hooks review path and remaining portal checks. Building a ZIP is not marketplace acceptance or publication.

Talk to Jay

Use /jay or /j to talk directly to Jay · AI Test Manager Agent. The MCP prompt router also accepts J. Skill names are lowercase; if a host does not resolve /J, use /j. Claude commonly prefixes these as /carbon:jay and /carbon:j; Codex may show a plugin-prefixed command or $jay in its skill picker. The plugin cannot reserve a global slash name in every host.

• /jay What should we test next? — discuss risks and the evidence we have.
• /j Test this project — run this edition's normal CARBON assessment.
• /jay Explain the last report — interpret recorded results without rerunning.
• /carbon remains the direct broad-testing command.

Jay replies in first person, in a frank, practical tone inspired by Jason Arbon. He is an AI persona, not Jason. With no clear testing target he asks for a URL, folder/repository, API, app/feature, or requirements. He does not create a demo unless asked. Questions do not automatically start tests, and fixes still need authorization. Lite keeps its existing capabilities; these are two conversational entrypoints, not two new testing workflows or an upgrade to Pro.

Jay's bundled icon appears in compatible command pickers and CARBON report views. Host chat avatars and image sizing are host-controlled; text-only chat uses “Jay · AI test manager” instead of a large image.

Jay quick checks

Jay gives changed code a quick second look by default with trusted host hooks. Existing disables remain off. Local idle checks: 2-minute delay, 3-minute runs, one worker, 3 runs and 10 reserved minutes per UTC day. No auto-fixes, installs or live-site requests. Source pattern matches are possible issues, not confirmed bugs. Jay reports category counts in a grayscale report with his icon. Optional Claude-only AI review and exact executable test commands require separate approval. The foreground coding agent uses its configured model access; idle checks make no model calls by default. Use carbon-background to pause, review, re-enable or disable everywhere. See behavior, controls and privacy. Unsupported hosts retain manual testing.

Explore CARBON Pro

Run /carbon-pro to open CARBON Pro features and pricing. In Claude Code or Cowork, the plugin-prefixed command is /carbon:carbon-pro. In Codex, choose carbon-pro from the plugin skill picker (or use $carbon-pro). MCP clients expose the carbon-pro prompt; the full runtime also accepts carbon_pro. This is a navigation shortcut available in every edition, not a paid testing workflow. It does not run tests, start checkout, install Pro, or expose customer downloads.

Optional usage analytics

Analytics records fixed command names, first activation, assessment lifecycle, report/settings use, edition and version. It never includes prompts, code, project paths, target URLs, screenshots, findings, credentials or a persistent user identifier. Enabled events go to Fathom at https://cdn.usefathom.com/; the service necessarily sees connection metadata such as IP address.

Claude editions retain default-enabled analytics. Codex editions default off; enabling requires the user's explicit ch