lr-ai-coding-agent

Execute arbitrary prompts via coding agent CLIs (Claude Code, GitHub Copilot, Gemini CLI). One interface for all coding agents — pass any prompt and get streaming output with tool use, cost tracking, and duration reporting.

Usage

Pass a prompt directly:

lr-ai-coding-agent -p "Refactor the auth module to use dependency injection"
lr-ai-coding-agent --prompt "Add unit tests for the UserService class"

Or pipe a prompt via stdin:

echo "Explain the architecture of this project" | lr-ai-coding-agent --stdin
cat my-prompt.txt | lr-ai-coding-agent --stdin

Target a specific directory:

lr-ai-coding-agent -p "Fix the failing tests" --working-dir /path/to/project

Options

Prerequisites

• .NET 10 SDK
• At least one coding agent CLI installed and available in PATH:
  • Claude Code (preferred) — requires ANTHROPIC_DEFAULT_SONNET_MODEL or ANTHROPIC_DEFAULT_OPUS_MODEL environment variable
  • GitHub Copilot CLI
  • Gemini CLI

Installation

dotnet tool install -g lr.Dotnet.AiCodingAgentTool

Or run directly without installing:

dotnet tool run lr-ai-coding-agent -- -p "your prompt here"

Building from source

git clone <repository-url>
cd lr_ai_coding_agent
dotnet build AiCodingAgent

To pack and publish the NuGet package:

./pack-cli.sh

This sets the package version from the current timestamp (YYYY.MM.DD.HHMM) and pushes to the configured NuGet feeds.