Multi-CLI Launcher

A powerful command-line tool for running multiple AI prompts in parallel using GitHub Copilot SDK. Execute multiple prompts concurrently, stream responses in real-time, and save consolidated results to markdown.

Features

• Parallel Execution: Run multiple prompts concurrently with configurable parallelism
• Real-time Streaming: View responses as they are generated in real-time
• Flexible Input: Support for inline prompts or prompts from a file
• Multi-line Prompts: Create complex prompts spanning multiple lines
• Model Selection: Choose your preferred AI model (default: Claude Opus 4.5)
• Timeout Control: Configure timeout per prompt to handle long-running tasks
• Skill Integration: Load custom skills from specified directories
• Markdown Output: Automatically save all results to a timestamped markdown file
• Error Handling: Graceful error handling with detailed error reporting

Requirements

• .NET 10.0 or later
• GitHub Copilot SDK access

Installation

1. Clone this repository:

git clone https://github.com/ManishJayaswal/multi-cli-launcher.git
cd multi-cli-launcher

2. Build the project:

dotnet build

3. Run the tool:

dotnet run -- [options] [prompts...]

Or build and run as a standalone executable:

dotnet publish -c Release
./bin/Release/net10.0/multi-cli-launcher [options] [prompts...]

Usage

Basic Syntax

multi-cli-launcher [options] [prompts...]

Command-Line Options

Option	Short	Description	Default
--model <name>	-m	AI model to use	Claude Opus 4.5
--file <path>	-f	Path to prompts file	-
--max-parallel <n>	-p	Maximum concurrent sessions	4
--output-dir <path>	-o	Output directory for results	./output
--timeout <minutes>	-t	Timeout per prompt in minutes	10
--skills <dir>	-s	Directory containing skill files (can be repeated)	-
--help	-h	Show help message	-

Examples

1. Run inline prompts

multi-cli-launcher "What is 2+2?" "Explain gravity in simple terms"

2. Run prompts from a file

multi-cli-launcher -f prompts.txt

3. Use a specific model

multi-cli-launcher -m "Claude Opus 4.5" -f prompts.txt

4. Limit parallelism and set custom output directory

multi-cli-launcher -p 2 -o ./results "Hello world" "Explain AI"

5. Handle long-running tasks with extended timeout

multi-cli-launcher -t 30 "Analyze this large dataset..."

6. Use custom skills

multi-cli-launcher -s ~/.copilot/skills "Use my custom skill"

7. Combine multiple options

multi-cli-launcher -f prompts.txt -m "Claude Opus 4.5" -p 3 -o ./output -t 15

Prompts File Format

Create a text file with your prompts. The format supports:

• Comments: Lines starting with # (except directives)
• Model Directive: Use #model: <model-name> at the top to override the default model
• Multi-line Prompts: Prompts can span multiple lines
• Prompt Delimiter: Use --- to separate prompts

Example Prompts File

#model: Claude Opus 4.5
# This is a comment

What is the capital of France?
---
# Multi-line prompt example
Explain the following concepts in simple terms:
- Recursion
- Big O notation
- Stack vs Heap
---
Write a short haiku about programming.

Prompts File Rules

1. Model Directive: Must appear before any prompts, prefixed with #model:
2. Comments: Lines starting with # (except #model:) are ignored
3. Delimiters: Use --- on its own line to separate prompts
4. Whitespace: Leading and trailing whitespace is trimmed from each prompt
5. Empty Prompts: Empty sections (containing only comments/whitespace) are skipped

Output Format

Console Output

The tool displays:

1. Configuration Summary: Shows model, parallelism, timeout, and prompt count
2. Real-time Streaming: Each prompt's response streams to the console as it's generated
3. Completion Status: Shows success/error count when all prompts complete

Example console output:

Model: Claude Opus 4.5
Max Parallel: 4
Timeout: 10 minutes
Output Directory: ./output
Total Prompts: 3
  [1] What is the capital of France?
  [2] Explain the following concepts in simple ter...
  [3] Write a short haiku about programming.

============================================================
## Prompt 1 (streaming...)
------------------------------------------------------------
What is the capital of France?
------------------------------------------------------------
### Response
The capital of France is Paris...
============================================================
[Prompt 1 completed]

Completed: 3 successful, 0 failed

Output saved to: ./output/Multi-CLI-Launcher-Output-2026-01-26-120530.md

Markdown Output

Results are saved to a timestamped markdown file in the output directory:

• Filename format: Multi-CLI-Launcher-Output-YYYY-MM-DD-HHmmss.md
• Contains all prompts and their responses
• Errors are included in code blocks
• Easy to share and review

Architecture

The tool is built with several key components:

• ArgumentParser: Parses command-line arguments into configuration
• PromptFileReader: Reads and parses prompts from files with directive support
• PromptExecutor: Executes prompts using GitHub Copilot SDK with streaming
• OutputCoordinator: Coordinates output from multiple parallel tasks using channels
• OutputConsumer: Consumes and formats output for console and markdown

Key Design Features

• Concurrency Control: Uses SemaphoreSlim to limit parallel executions
• Streaming Support: Real-time streaming of responses to console
• Channel-based Coordination: Prevents output interleaving using .NET Channels
• First-Complete-First-Print: Displays results as they complete, not in order
• Independent Sessions: Each prompt runs in its own isolated session

Testing

The project includes comprehensive unit tests:

# Run all tests
dotnet test

# Run tests with verbose output
dotnet test -v n

# Run specific test file
dotnet test --filter "FullyQualifiedName~ArgumentParserTests"

Test coverage includes:

• Argument parsing
• Prompt file reading
• Output coordination
• Output consumption

Contributing

Contributions are welcome! Please feel free to submit issues or pull requests.

License

[Specify your license here]

Author

Manish Jayaswal

Acknowledgments

Built with the GitHub Copilot SDK