Coverctl

You are running an MCP server designed to expose coverage tooling to AI agents and automated workflows. It provides a programmable interface for running coverage checks, generating reports, and recording history, enabling automated policy enforcement and insights across your multi-language projects.

How to use

You start the MCP server for your project by launching the CLI in your repository. The server runs as a local, stdio-based endpoint that AI agents can interact with through standard input and output. With the server running, you can trigger coverage checks, generate reports, and record history, all through the MCP protocol inside your development environment or CI integrations.

How to install

Prerequisites: You need Go installed on your system to install the MCP CLI. Ensure you have a compatible Go toolchain and access to the Go command line. You will install the MCP client/binary and then start the MCP server in your project.

Step by step commands you will run in your terminal:

# Install the MCP client binary
go install github.com/felixgeelhaar/coverctl@latest

# Navigate to your project directory
cd your-project

# Start the MCP server for the current project
coverctl mcp serve

Additional sections

Configuration and usage patterns described here assume you use the MCP server to integrate coverage tooling with AI agents like Claude, or with automated CI workflows. You can run the server in the same directory as your project to expose its endpoints over STDIO for the agents to consume.

Claude Desktop integration: You can configure Claude to communicate with the MCP server so Claude can perform coverage checks, analyze reports, and provide recommendations based on your project data. A typical local configuration maps a named MCP server to the coverctl command and its arguments.

Example Claude Desktop configuration for a local project:

```json
{
  "mcpServers": {
    "coverctl": {
      "command": "coverctl",
      "args": ["mcp", "serve"],
      "cwd": "/path/to/your/go/project"
    }
  }
}

This enables Claude to run coverage checks, analyze reports, and provide recommendations based on your project’s coverage data.

## Notes

The MCP server is designed to be used locally. When you run it from your project directory, it reads your repository configuration if available and exposes a protocol surface that AI agents can query or command. Use this approach to automate coverage workflows, surface actionable insights, and keep your project aligned with your coverage goals.

## Available tools

### check

Run coverage tests and enforce policy thresholds for the detected domains.

### report

Analyze an existing coverage profile against the configured domains and thresholds.

### record

Record current coverage to history to track trends over time.

### trend

Show coverage trends over time using historical data.

### debt

Show coverage debt for domains/files below target and estimate remediation effort.

### suggest

Suggest optimal coverage thresholds and allow applying them to the configuration.

### mcp serve

Start the MCP server to expose tooling capabilities to AI agents via STDIO.