coverit

Your code, covered. AI-powered test quality.

What is coverit?

coverit is an AI-powered test quality platform. It scans and analyzes your entire codebase, identifies coverage gaps across 5 quality dimensions, generates tests to fill them, and tracks your score over time — all through a single persistent manifest: coverit.json.

Four commands. That's it.

Installation

1. Claude Code (Plugin)

Install the plugin for MCP tools, slash commands, and CLAUDE.md instructions — all in one step:

/plugin marketplace add devness-com/coverit
/plugin install coverit@coverit

This gives you:

• MCP tools (coverit_scan, coverit_cover, coverit_fix, coverit_status, etc.)
• Slash commands (/coverit:scan, /coverit:cover, /coverit:fix, /coverit:status)

2. Other AI Tools (MCP)

For Cursor, Windsurf, VS Code, Gemini CLI, Zed, Cline, Codex, Goose, and more:

npx @devness/coverit

This scans for AI tools on your machine, lets you choose which to configure, and sets up the MCP server interactively. Manage your installation anytime:

npx @devness/coverit              # Install / reconfigure
npx @devness/coverit --remove     # Uninstall from configured tools
npx @devness/coverit --status     # Show configuration status

3. CLI (Standalone)

Use directly via npx in any project — no MCP or plugin required:

npx @devness/coverit scan      # AI scans & analyzes your codebase → creates coverit.json
npx @devness/coverit cover     # AI writes tests for gaps → updates your score
npx @devness/coverit fix       # Fix failing tests via AI → updates your score
npx @devness/coverit status    # Show your quality dashboard

How It Works

scan → coverit.json → cover/fix → updated coverit.json → status

Command	What happens
scan	AI scans and analyzes your codebase with tool access (read files, search code, run commands). Discovers modules, maps existing tests, classifies complexity, identifies user journeys and API contracts. Produces coverit.json.
cover	AI reads gaps from coverit.json, writes test files for each module, runs them, fixes failures, and updates the quality score.
fix	Runs existing tests, sends failures to AI for fixing, rescans and updates the quality score. Unlike cover, doesn't write new tests.
status	Instantly displays your quality dashboard from coverit.json. No AI, no scanning.

The Manifest: coverit.json

coverit.json is the single source of truth. It's git-tracked, persistent, and contains:

• Project info — framework, language, test runner, file/line counts
• Modules — each source directory with complexity, file counts, and test coverage (current vs expected)
• Quality score — 0-100, weighted across 5 dimensions (ISO 25010)
• Gaps — what's missing, ranked by severity
• Journeys — user flows that need E2E coverage
• Contracts — API endpoints that need contract tests

CLI Reference

Command	Description
coverit scan [path]	AI scans and analyzes codebase → creates coverit.json
coverit cover [path]	AI generates tests from gaps and updates score
coverit fix [path]	Fix failing tests via AI, update score
coverit status [path]	Show quality dashboard from coverit.json
coverit clear [path]	Delete coverit.json and .coverit/ for a fresh start

Cover Options

Option	Description
--modules <paths>	Only cover specific modules (comma-separated)

Fix Options

Option	Description
--modules <paths>	Only fix tests for specific modules (comma-separated)

MCP Tools

Available in all AI tools after installation (plugin or npx):

Tool	Description
coverit_scan	AI scans and analyzes codebase → coverit.json
coverit_cover	AI generates tests from gaps → updates score
coverit_fix	Fix failing tests via AI → update score
coverit_status	Show quality dashboard (instant, no AI)
coverit_clear	Delete coverit.json and .coverit/
coverit_backup	Export coverit.json as JSON
coverit_restore	Import coverit.json from backup

Quality Dimensions

coverit measures quality across 5 dimensions mapped to ISO 25010:

Dimension	What it checks
Functionality	Unit, integration, API, E2E, and contract test coverage
Security	OWASP Top 10, injection, auth bypass, data exposure
Stability	Error handling, edge cases, resource cleanup
Conformance	Naming conventions, patterns, architectural rules
Regression	Baseline test results, breaking changes

Testing Diamond

coverit follows the Testing Diamond strategy (not the outdated Testing Pyramid):

Test Type	Target %	Purpose
Integration	~50%	Module boundaries, real dependencies
Unit	~20%	Pure functions, algorithms, edge cases
API	~15%	HTTP endpoints, request/response contracts
E2E	~10%	Critical user journeys
Contract	~5%	API schema validation

Supported Frameworks

Category	Frameworks
API	Hono, Express, NestJS, Fastify
Frontend	React, Next.js
Mobile	React Native, Expo
Desktop	Tauri, Electron
Test Runners	Vitest, Jest, Playwright, Detox, Pytest
Package Managers	Bun, pnpm, npm, yarn

Architecture

┌──────────────┐
│  CLI / MCP   │  ← Entry points (4 commands)
└──────┬───────┘
       │
       ├── scan ────→ AI scans & analyzes ──→ coverit.json
       │
       ├── cover ───→ AI reads gaps ──→ writes tests ──→ runs ──→ updates coverit.json
       │
       ├── fix ─────→ runs tests ──→ AI fixes failures ──→ updates coverit.json
       │
       └── status ──→ reads coverit.json ──→ dashboard

src/
├── ai/           AI providers (Claude, Gemini, Codex, Anthropic, OpenAI, Ollama)
├── scale/        Codebase scanner + manifest writer
├── cover/        Test generation pipeline
├── fix/          Test fix pipeline
├── measure/      Test scanner, scorer, dashboard
├── scoring/      Score engine, weights, thresholds
├── schema/       coverit.json types + defaults
├── types/        Core type definitions
├── utils/        Logger, framework detector
├── mcp/          MCP server (7 tools)
└── cli/          CLI (5 commands)

See docs/architecture.md for details.

Contributing

Contributing guide · Report an issue

License

AGPL-3.0 · devness-com/coverit