Skip to content

CovLoupe

CovLoupe logo

An MCP server, command line utility, and library for Ruby SimpleCov test coverage analysis.

Documentation Website

Gem Version Documentation

What is cov-loupe?

cov-loupe makes SimpleCov coverage data queryable and actionable through three interfaces:

  • MCP server - Lets AI assistants analyze your coverage
  • CLI - Fast command-line coverage reports and queries
  • Ruby library - Programmatic API for custom tooling

Works with any SimpleCov-generated .resultset.json file—no runtime dependency on your test suite.

Key Features

  • ✅ Multiple interfaces - MCP server, CLI, and Ruby API
  • Annotated source code - -s full|uncovered / --source full|uncovered with -c N / --context-lines N for context lines
  • ✅ Staleness detection - Identify outdated coverage (missing files, timestamp mismatches, line count changes)
  • ✅ Multi-suite support - Automatic merging of multiple test suites (RSpec + Cucumber, etc.)
  • ✅ Flexible path resolution - Works with absolute or relative paths
  • ✅ Comprehensive error handling - Context-aware messages for each mode

Practical Use Cases

  • Query coverage data from AI assistants, e.g.:
  • "Using cov-loupe, analyze test coverage data and write a report to a markdown file containing a free text analysis of each issue and then two tables, one sorted in descending order of urgency, the other in ascending order of level of effort."
  • "Using cov-loupe, generate a table of directories and their average coverage rates, in ascending order of coverage."
  • Find files with the lowest coverage
  • Investigate specific files or directories
  • Generate CI/CD coverage reports
  • Create custom pass/fail predicates for scripts and CI - use the library API or CLI JSON output to implement arbitrarily complex coverage rules beyond simple thresholds (e.g., require higher coverage for critical paths, exempt test utilities, track coverage trends)

Quick Start

Installation

gem install cov-loupe

Upgrading

If you are upgrading from a previous version, please refer to the Migration Guides.

Generate Coverage Data

# Run your tests with SimpleCov enabled
bundle exec rspec  # or your test command

# Verify coverage was generated
ls -l coverage/.resultset.json

Basic Usage

CLI - View Coverage Table: 

cov-loupe

CLI - Check Specific File: 

cov-loupe summary lib/cov_loupe/model.rb
cov-loupe uncovered lib/cov_loupe/cli.rb

CLI - Find the Project Homepage Fast: Run cov-loupe -h and the banner's second line shows the repository URL. Some terminal applications (e.g. iTerm2) will enable direct clicking the link using modifier keys such as Cmd or Alt. 

Usage:      cov-loupe [options] [subcommand] [args]
Repository: https://github.com/keithrbennett/cov-loupe  # <--- Project URL ---

Ruby Library: 

require "cov_loupe"

model = CovLoupe::CoverageModel.new
list_result = model.list
files = list_result["files"]
# => [{ "file" => "lib/cov_loupe/model.rb", "covered" => 114, "total" => 118, "percentage" => 96.61, "stale" => false }, ...]

summary = model.summary_for("lib/cov_loupe/model.rb")
# => { "file" => "lib/cov_loupe/model.rb", "summary" => { "covered" => 114, "total" => 118, "percentage" => 96.61 } }

For advanced use cases, multiple models can each have their own data source and log file. See Library API for details.

MCP Server: See MCP Integration Guide for AI assistant setup.

Multi-Suite Coverage

Projects with multiple test suites (RSpec + Cucumber, etc.) are automatically merged. See Multi-Suite Coverage Merging for details and current limitations.

Documentation Index

Full documentation is available at https://keithrbennett.github.io/cov-loupe/.

User Guides:

Special Topics & Prompts:

Developer Docs:

Project Docs & Examples:

Requirements

  • Ruby >= 3.2 (required by mcp gem dependency)
  • SimpleCov-generated .resultset.json file
  • simplecov gem >= 0.21

Configuring the Resultset

cov-loupe automatically searches for .resultset.json in standard locations (coverage/.resultset.json, .resultset.json, tmp/.resultset.json). For non-standard locations:

# Command-line option (highest priority) - use -r or --resultset
cov-loupe -r /path/to/your/coverage

# Environment variable (project-wide default)
export COV_LOUPE_OPTS="-r /path/to/your/coverage"

# MCP server configuration
# Add to your MCP client config (used as defaults for MCP tools):
# "args": ["-r", "/path/to/your/coverage"]

MCP precedence: For MCP tool calls, per-request JSON parameters win over the CLI args used to start the server (including COV_LOUPE_OPTS). If neither is provided, built-in defaults are used (root: '.', raise_on_stale: false, etc.). Coverage data is cached globally and automatically reloaded when the resultset file changes.

See CLI Usage Guide for complete details.

Common Workflows

Find Coverage Gaps

# Files with worst coverage
cov-loupe -o d list           # -o = --sort-order, d = descending (worst at end)
cov-loupe list | less         # display table in pager, best files first (worst at end)
cov-loupe list | head -10     # truncate the table

# Filter to specific patterns (see COV_LOUPE_OPTS best practice below)
cov-loupe -g "lib/cov_loupe/tools/**/*.rb" list  # -g = --tracked-globs

# Export for analysis
cov-loupe -fJ list > coverage-report.json

Best Practice: Match SimpleCov Configuration

For accurate coverage tracking and validation, set COV_LOUPE_OPTS to match your SimpleCov track_files patterns:

# In spec_helper.rb or rails_helper.rb
SimpleCov.start do
  add_filter '/spec/'
  track_files 'lib/**/*.rb'
  track_files 'app/**/*.rb'
end

# In your shell config (.bashrc, .zshrc, etc.)
export COV_LOUPE_OPTS="--tracked-globs lib/**/*.rb,app/**/*.rb"

This ensures list and totals output matches SimpleCov's scope and missing_tracked_files (in list) / missing_from_coverage (in totals) report meaningful gaps.

Note: By default, --tracked-globs is empty (shows all files in the resultset). This prevents silently hiding coverage data that doesn't match assumed patterns.

Working with JSON Output

The -fJ flag enables programmatic processing of coverage data using command-line JSON tools.

Using jq: 

# Filter files below 80% coverage
cov-loupe -fJ list | jq '.files[] | select(.percentage < 80)'

Using Ruby one-liners: 

# Count files below threshold
cov-loupe -fJ list | ruby -r json -e '
  puts JSON.parse($stdin.read)["files"].count { |f| f["percentage"] < 80 }
'

Using rexe:

rexe is a Ruby gem that enables shorter Ruby command lines by providing command-line options for input and output formats, plus other conveniences. It eliminates the need for explicit JSON parsing and formatting code.

Install: gem install rexe

# Filter files below 80% coverage with pretty-printed JSON output
cov-loupe -fJ list | rexe -ij -mb -oJ 'self["files"].select { |f| f["percentage"] < 80 }'

# Count files below threshold
cov-loupe -fJ list | rexe -ij -mb -op 'self["files"].count { |f| f["percentage"] < 80 }'

# Human-readable output with AmazingPrint
cov-loupe -fJ list | rexe -ij -mb -oa 'self["files"].first(3)'

With rexe's -ij -mb options, self automatically becomes the parsed JSON object. The same holds true for JSON output -- using -oJ produces pretty-printed JSON without explicit formatting calls. Rexe also supports YAML input/output (-iy, -oy) and AmazingPrint output (-oa) for human consumption.

When Coverage Rows Are Skipped

If a file in the resultset is missing on disk or has corrupt data, the CLI now logs and warns after rendering the report so operators immediately see that totals may be incomplete. Example table output:

$ cov-loupe list
┌─────────────────────────────┬─────────┬─────────┬────────────┬───────┐
│ File                        │ Covered │ Total   │ % Covered  │ Stale │
├─────────────────────────────┼─────────┼─────────┼────────────┼───────┤
│ lib/foo.rb                  │       2 │       3 │     66.67% │       │
│ lib/bar.rb                  │       1 │       3 │     33.33% │       │
└─────────────────────────────┴─────────┴─────────┴────────────┴───────┘

WARNING: 1 coverage row skipped due to errors:
  - lib/deleted.rb: No coverage data found for file: lib/deleted.rb
Run again with --raise-on-stale to exit when rows are skipped.

Pretty JSON (-fJ) reports still emit valid JSON to stdout; the warning continues to be printed on stderr:

$ cov-loupe -fJ list
{
  "files": [
    { "file": "lib/foo.rb", "covered": 2, "total": 3, "percentage": 66.67, "stale": null },
    { "file": "lib/bar.rb", "covered": 1, "total": 3, "percentage": 33.33, "stale": null }
  ],
  "counts": { "total": 2, "ok": 2, "stale": 0 }
}

WARNING: 1 coverage row skipped due to errors:
  - lib/deleted.rb: No coverage data found for file: lib/deleted.rb
Run again with --raise-on-stale to exit when rows are skipped.

Use --raise-on-stale true (or -S true) to turn these warnings into hard failures for CI pipelines.

Run rexe -h to see all available options, or visit the rexe project page for more examples.

For comprehensive JSON processing examples, see user/EXAMPLES.md.

CI/CD Integration

# Fail build if coverage is stale (--raise-on-stale or -S)
cov-loupe --raise-on-stale true list || exit 1

# Generate coverage report artifact
cov-loupe -fJ list > artifacts/coverage.json

Investigate Specific Files

# Quick summary
cov-loupe summary lib/cov_loupe/model.rb

# See uncovered lines
cov-loupe uncovered lib/cov_loupe/cli.rb

# View in context
cov-loupe -s u -c 3 uncovered lib/cov_loupe/cli.rb  # -s = --source (u = uncovered), -c = --context-lines

# Detailed hit counts
cov-loupe detailed lib/cov_loupe/util.rb

# Project totals
cov-loupe totals
cov-loupe -fJ totals

Boolean CLI Options

Boolean flags such as --color (short: -C) and --raise-on-stale (short: -S) require explicit boolean arguments. Recognized literals:
yes no
y n
true false
t f
on off
+ -
1 0

Each row lists the equivalent true token (left) and false token (right).

cov-loupe --color false        # disable ANSI colors explicitly
cov-loupe -C false             # short form
cov-loupe --raise-on-stale yes # enforce stale coverage failures

Commands and Tools

CLI Subcommands: list, summary, uncovered, detailed, raw, totals, validate, version

MCP Tools: coverage_summary_tool, coverage_detailed_tool, coverage_raw_tool, uncovered_lines_tool, list_tool, coverage_totals_tool, coverage_table_tool, validate_tool, help_tool, version_tool

📖 See also: - CLI Usage Guide - Complete command-line reference - MCP Integration Guide - MCP tools documentation

Troubleshooting

  • "command not found" - See Installation Guide
  • "cannot load such file -- mcp" - Requires Ruby >= 3.2. Verify: ruby -v
  • "Could not find .resultset.json" - Ensure SimpleCov is configured in your test suite, then run tests to generate coverage. See the Configuring the Resultset section for more details.
  • MCP server won't connect - Check PATH and Ruby version in MCP Troubleshooting
  • RVM in sandboxed environments (macOS) - RVM requires /bin/ps which may be blocked by sandbox restrictions. Use rbenv or chruby instead.

For more detailed help, see the full Troubleshooting Guide.

Development

# Clone and setup
git clone https://github.com/keithrbennett/cov-loupe.git
cd cov-loupe
bundle install

# Run tests
bundle exec rspec

# Test locally
bundle exec exe/cov-loupe

# Build and install
gem build cov-loupe.gemspec
gem install cov-loupe-*.gem

See dev/DEVELOPMENT.md for more.

SimpleCov Dependency

cov-loupe declares a runtime dependency on simplecov (>= 0.21) to support multi-suite merging using SimpleCov's combine helpers. The dependency is lazy-loaded only when needed, ensuring fast startup for single-suite projects.

Contributing

Contributions are welcome! Please:

  1. Fork the repository
  2. Create a feature branch
  3. Add tests for new functionality
  4. Ensure all tests pass (bundle exec rspec)
  5. Submit a pull request

License

MIT License - see LICENSE file for details.

  • GitHub: https://github.com/keithrbennett/cov-loupe
  • RubyGems: https://rubygems.org/gems/cov-loupe
  • Issues: https://github.com/keithrbennett/cov-loupe/issues
  • Changelog: RELEASE_NOTES.md

Next Steps

📦 Install: gem install cov-loupe

📖 Read: CLI Usage Guide | MCP Integration

🐛 Report issues: GitHub Issues

⭐ Star the repo if you find it useful!