5 Example Plugins for Testing and QA in Claude: Complete Guide

Claude Code supports testing and QA workflows through .claude-plugin bundles such as Meticulous, Agentiqa, QA‑Orchestra, TestDino, and 42Crunch, each exposing MCP tools for visual regression, AI-driven exploration, and automated security scanning.

The anthropics/claude-plugins-community repository maintains a marketplace of community-built plugins that extend Claude Code with specialized quality-assurance capabilities. These plugins leverage the Model-Centered Programming (MCP) protocol to connect Claude with external testing services, allowing you to execute test runs, debug failures, and generate QA reports using natural language commands.

Meticulous: Visual Regression Testing

Meticulous provides visual-regression testing capabilities that let you review test runs, investigate replays, and debug visual diffs directly within Claude Code. The plugin connects to a hosted Meticulous MCP server and exposes slash commands for managing visual test campaigns.

In .claude-plugin/marketplace.json at lines 100–111, the Meticulous entry is flagged with "category": "testing" and declares the MCP tools necessary for visual comparison workflows. The plugin manifest defines how Claude should communicate with the Meticulous backend to fetch screenshots and diff data.

To install and use Meticulous:

/plugin marketplace add meticulous
/meticulous list-runs
/meticulous diff <run-id>

The /meticulous list-runs command retrieves recent visual-regression runs, while /meticulous diff opens an interactive diff view for debugging specific test failures.

Agentiqa: AI-Powered QA Agent

Agentiqa functions as an autonomous QA agent that visually explores web or mobile applications without requiring traditional test scripts. It reports bugs with severity ratings, confidence scores, and step-by-step reproduction instructions generated through visual analysis.

The plugin definition in .claude-plugin/marketplace.json at lines 768–775 marks Agentiqa as "category": "testing" and configures the MCP tools needed for visual session management. Unlike assertion-based frameworks, Agentiqa uses computer vision to detect UI anomalies and behavioral inconsistencies.

Usage example:

/plugin marketplace add agentiqa
/agentiqa test "https://my-app.com"

This initiates an exploratory testing session where Claude delegates visual navigation to the Agentiqa MCP server, then summarizes findings with structured bug reports.

QA-Orchestra: Multi-Agent QA Suite

QA-Orchestra bundles ten specialized QA agents into a single plugin, including the functional-reviewer, test-scenario-designer, smart-test-selector, and bug-reporter. This suite answers pull-request-related QA questions and outputs structured Markdown reports.

According to .claude-plugin/marketplace.json at lines 16052–16056, the QA-Orchestra entry groups multiple agent definitions under one namespace while maintaining the "category": "testing" classification. Each sub-agent handles a specific phase of the quality-assurance lifecycle, from initial scenario generation to final bug triage.

Typical workflow:

/plugin marketplace add qa-orchestra
/qa-orchestra functional-reviewer
/qa-orchestra test-scenario-designer
/qa-orchestra bug-reporter

These commands invoke distinct agents within the suite, allowing you to chain functional analysis with test-case generation and automated reporting in a single conversation.

TestDino: Playwright and Manual Testing

TestDino offers remote MCP-based integration with the TestDino platform, supporting Playwright-based CI tests, flaky test debugging, and manual test case management. The plugin is explicitly tagged for QA workflows in its manifest file.

In testdino/.claude-plugin/plugin.json at lines 18–20, the plugin metadata includes a "qa" tag alongside its core definition, signaling its testing focus. The accompanying README at lines 8–15 describes capabilities for running Playwright specs, maintaining manual test suites, and conducting automated audits.

Installation and commands:

/plugin marketplace add testdino-hq/TestDino-Plugins
/testdino-runs list
/testdino-runs debug "visual.spec.js"
/testdino-manual-tests create "Checkout flow"

The /testdino-runs namespace handles CI execution history and debugging, while /testdino-manual-tests manages human-verified test cases separate from automated suites.

42Crunch: API Security Testing

42Crunch API Security Testing performs automated security scans including OpenAPI static analysis, OWASP API risk detection, and live conformance testing. Though security-focused, it resides in the "category": "testing" classification alongside functional QA tools.

The marketplace entry in .claude-plugin/marketplace.json at lines 54–61 defines the plugin’s MCP tools for auditing API specifications and executing dynamic scans against live endpoints. This allows Claude to validate contract security without leaving the chat interface.

Security testing commands:

/plugin marketplace add 42crunch-api-security-testing
/42crunch audit-openapi "./openapi.yaml"
/42crunch live-scan "https://api.myservice.com"

The /42crunch audit-openapi command performs static analysis on local specification files, while /42crunch live-scan probes running services for runtime vulnerabilities.

How Testing Plugins Work in Claude

Testing plugins in the anthropics/claude-plugins-community repository follow a standardized architecture that enables discovery and composition.

Each plugin requires:

• Plugin Manifest (.claude-plugin/plugin.json) – Declares the plugin name, description, required configuration (API keys, endpoints), and category tags.
• Skill Definitions (SKILL.md plus optional references/ and pipelines/ directories) – Provide natural-language instructions Claude uses to invoke MCP tools correctly.
• Category Tag – The "category": "testing" label in marketplace.json enables filtering in the marketplace UI.
• MCP Integration – Either hosted endpoints (Meticulous, TestDino) or local services implementing the declared tool schemas.

When you execute /plugin marketplace add <name>, Claude loads the skill definitions into the current session, making testing capabilities available through namespaced slash commands. Because all testing plugins use MCP, they can be composed in sequences—such as running a Meticulous visual diff followed by an Agentiqa exploration—without additional integration code.

Summary

• Meticulous handles visual-regression testing via a hosted MCP server, with commands for listing runs and debugging diffs.
• Agentiqa provides AI-driven exploratory testing of web and mobile apps, generating bug reports with confidence scores.
• QA-Orchestra offers ten specialized agents for functional review, scenario design, and structured bug reporting.
• TestDino integrates Playwright CI execution and manual test management through remote MCP connections.
• 42Crunch delivers automated API security scanning, combining static OpenAPI audits with live conformance testing.
• All testing plugins are categorized in .claude-plugin/marketplace.json and expose capabilities through MCP-based slash commands.

Frequently Asked Questions

How do I install testing plugins in Claude Code?

Run the /plugin marketplace add <plugin-name> command in Claude Code, where <plugin-name> corresponds to the "name" field in .claude-plugin/marketplace.json. After installation, the plugin's slash commands become available immediately in your current session. For example, plugin marketplace add meticulous loads the Meticulous visual-regression tools.

What is the difference between MCP tools and Claude plugins?

MCP (Model-Centered Programming) is the underlying protocol that defines how Claude communicates with external services. A Claude plugin is a .claude-plugin bundle containing a plugin.json manifest and SKILL.md definitions that map natural language to MCP tool calls. Plugins package MCP capabilities into discoverable, categorized components with specific versioning and configuration requirements.

Can I combine multiple testing plugins in one workflow?

Yes. Because all testing plugins in the anthropics/claude-plugins-community repository implement the MCP protocol, Claude can orchestrate them together without glue code. You can run a Meticulous visual diff, then invoke Agentiqa to verify UI behavior, and finally generate a QA-Orchestra report—all within the same conversation context.

Where are testing plugin definitions stored in the repository?

Plugin metadata resides in .claude-plugin/marketplace.json at the repository root, with individual plugin manifests housed in subdirectories like testdino/.claude-plugin/plugin.json. The marketplace file uses line-specific entries (e.g., lines 100–111 for Meticulous, lines 768–775 for Agentiqa) to declare category tags and MCP server configurations.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}