A simple command-line interface for sending prompts to LLM APIs via stdin. Supports OpenRouter, Ollama, and LM Studio. Perfect for piping data from other commands.

This CLI tool reads input from stdin and sends it to various LLM providers. It's designed to work seamlessly with Unix pipes, making it easy to process command output through AI models. Supports both cloud (OpenRouter) and local (Ollama, LM Studio) LLM providers.

• 🚀 Piped Input: Read from stdin for seamless integration with other commands
• 🔒 Environment-Based Configuration: All settings via environment variables
• 🛡️ Robust Error Handling: Clear error messages and proper HTTP handling
• 📡 Streaming Support: Real-time streaming responses (SSE for OpenRouter/LM Studio, newline-delimited JSON for Ollama)
• 🔍 Verbose/Debug Mode: Detailed logging for troubleshooting
• 🏠 Local LLM Support: Works with Ollama and LM Studio for offline/local AI processing
• ☁️ Cloud LLM Support: Works with OpenRouter for cloud-based AI models

• Go 1.21 or later
• For OpenRouter: An OpenRouter API key (get one from OpenRouter)
• For Ollama: Install and run Ollama locally
• For LM Studio: Install and run LM Studio with the local server enabled

The project includes a Makefile with cross-platform build support. All binaries are built in the bin/ directory.

# Navigate to the project directory
cd openrouter-cli

# Build for your current platform
make build

# Or use go directly
go build -o bin/openrouter-cli .

Build for all supported platforms:

# Build for all platforms (Linux, macOS, Windows)
make all

Build for a specific platform:

# Linux AMD64
make linux-amd64

# Linux ARM64
make linux-arm64

# macOS Intel (AMD64)
make darwin-amd64

# macOS Apple Silicon (ARM64)
make darwin-arm64

# Windows 64-bit
make windows-amd64

# Windows 32-bit
make windows-386

All binaries will be placed in the bin/ directory with platform-specific suffixes:

• Linux: openrouter-cli-linux-amd64 or openrouter-cli-linux-arm64
• macOS: openrouter-cli-darwin-amd64 or openrouter-cli-darwin-arm64
• Windows: openrouter-cli-windows-amd64.exe or openrouter-cli-windows-386.exe

After building, you can install the binary:

# Install to system directory (requires sudo)
make install

# Install to user's local bin directory (no sudo required)
make install-user

# Clean build artifacts
make clean

# Format code
make fmt

# Run tests
make test

# Run tests with coverage
make test-coverage

# Show help
make help

Note: In the examples below, ./openrouter-cli refers to the binary. If you built it with make build, use bin/openrouter-cli. If you installed it with make install or make install-user, you can use openrouter-cli directly (assuming it's in your PATH).

The tool reads from stdin. Choose your provider:

Linux/macOS:

# Set your API key and provider
export OPENROUTER_API_KEY="your-api-key-here"
export LLM_PROVIDER="openrouter"

# Pipe command output to OpenRouter
ps aux | ./openrouter-cli

# Pipe file content
cat file.txt | ./openrouter-cli

Windows (PowerShell):

# Set your API key and provider
$env:OPENROUTER_API_KEY="your-api-key-here"
$env:LLM_PROVIDER="openrouter"

# Pipe command output to OpenRouter
Get-Process | .\openrouter-cli-windows-amd64.exe

# Pipe file content
Get-Content file.txt | .\openrouter-cli-windows-amd64.exe

Windows (Command Prompt):

REM Set your API key and provider
set OPENROUTER_API_KEY=your-api-key-here
set LLM_PROVIDER=openrouter

REM Pipe command output to OpenRouter
tasklist | openrouter-cli-windows-amd64.exe

REM Pipe file content
type file.txt | openrouter-cli-windows-amd64.exe

Linux/macOS:

# Set provider to Ollama (no API key needed)
export LLM_PROVIDER="ollama"
export LLM_MODEL="llama2"  # or any model you have installed

# Make sure Ollama is running, then pipe command output
ps aux | ./openrouter-cli

Windows (PowerShell):

# Set provider to Ollama (no API key needed)
$env:LLM_PROVIDER="ollama"
$env:LLM_MODEL="llama2"

# Make sure Ollama is running, then pipe command output
Get-Process | .\openrouter-cli-windows-amd64.exe

Linux/macOS:

# Set provider to LM Studio (no API key needed)
export LLM_PROVIDER="lmstudio"
export LLM_MODEL="local-model"  # or the model name in LM Studio

# Make sure LM Studio server is running, then pipe command output
ps aux | ./openrouter-cli

Windows (PowerShell):

# Set provider to LM Studio (no API key needed)
$env:LLM_PROVIDER="lmstudio"
$env:LLM_MODEL="local-model"

# Make sure LM Studio server is running, then pipe command output
Get-Process | .\openrouter-cli-windows-amd64.exe

The tool uses the following environment variables:

• LLM_PROVIDER (optional): Provider to use - openrouter, ollama, or lmstudio (default: openrouter)
• OPENROUTER_API_KEY (required for OpenRouter): Your OpenRouter API key
• LLM_MODEL (optional): Model to use
  • OpenRouter default: openai/gpt-oss-20b:free
  • Ollama default: llama2
  • LM Studio default: local-model
• LLM_PRE_PROMPT (optional): Text to prepend to the stdin input
• LLM_STREAM (optional): Enable streaming responses. Set to 1, true, yes, or on to enable
• LLM_VERBOSE (optional): Enable verbose/debug logging. Set to 1, true, yes, or on to enable
• LLM_TIMEOUT (optional): Request timeout in seconds (default: 60 for non-streaming, 300 for streaming). Useful for slow local LLMs like LM Studio
• OLLAMA_URL (optional): Ollama API URL (default: http://localhost:11434/api/chat)
• LM_STUDIO_URL (optional): LM Studio API URL (default: http://localhost:1234/v1/chat/completions)

Linux/macOS (Bash/Zsh):

export OPENROUTER_API_KEY="your-api-key-here"
export LLM_PROVIDER="openrouter"
export LLM_MODEL="openai/gpt-4"

Windows Command Prompt (cmd.exe):

set OPENROUTER_API_KEY=your-api-key-here
set LLM_PROVIDER=openrouter
set LLM_MODEL=openai/gpt-4

Windows PowerShell:

$env:OPENROUTER_API_KEY="your-api-key-here"
$env:LLM_PROVIDER="openrouter"
$env:LLM_MODEL="openai/gpt-4"

Note: Environment variables set in Command Prompt or PowerShell are session-specific. To make them persistent, use System Properties → Environment Variables, or set them in your shell profile.

# Analyze process list
ps aux | ./openrouter-cli

# Analyze log file
tail -n 100 app.log | ./openrouter-cli

# Analyze command output
df -h | ./openrouter-cli

# OpenRouter
export OPENROUTER_API_KEY="your-api-key"
export LLM_PROVIDER="openrouter"
export LLM_MODEL="openai/gpt-4"
ps aux | ./openrouter-cli

# Ollama
export LLM_PROVIDER="ollama"
export LLM_MODEL="llama3.2"
ps aux | ./openrouter-cli

# LM Studio
export LLM_PROVIDER="lmstudio"
export LLM_MODEL="mistral-7b-instruct"
ps aux | ./openrouter-cli

# Works with any provider
export LLM_PROVIDER="ollama"  # or "openrouter" or "lmstudio"
export OPENROUTER_API_KEY="your-api-key"  # only needed for openrouter
export LLM_PRE_PROMPT="Analyze the following process list and identify any suspicious processes:"

ps aux | ./openrouter-cli

export LLM_PROVIDER="ollama"  # or "openrouter" or "lmstudio"
export OPENROUTER_API_KEY="your-api-key"  # only needed for openrouter
export LLM_STREAM="true"

echo "Write a short story about a robot" | ./openrouter-cli

Streaming mode outputs responses in real-time as they're generated, providing a better user experience for longer responses.

export LLM_PROVIDER="ollama"  # or "openrouter" or "lmstudio"
export OPENROUTER_API_KEY="your-api-key"  # only needed for openrouter
export LLM_VERBOSE="true"

echo "Hello world" | ./openrouter-cli

Verbose mode provides detailed logging including:

• Request/response details
• HTTP status codes and headers
• Input/output sizes
• Streaming chunk information
• Error details

# OpenRouter example
export LLM_PROVIDER="openrouter"
export OPENROUTER_API_KEY="your-api-key"
export LLM_MODEL="anthropic/claude-3-opus"
export LLM_PRE_PROMPT="Summarize the following:"
export LLM_STREAM="true"
export LLM_VERBOSE="true"
cat document.txt | ./openrouter-cli

# Ollama example
export LLM_PROVIDER="ollama"
export LLM_MODEL="llama3.2"
export LLM_PRE_PROMPT="Summarize the following:"
export LLM_STREAM="true"
export LLM_VERBOSE="true"
cat document.txt | ./openrouter-cli

• Requires an API key
• Supports all models available on OpenRouter
• Uses OpenAI-compatible API format
• Streaming uses Server-Sent Events (SSE)

• No API key required
• Make sure Ollama is running: ollama serve
• Install models: ollama pull llama2 (or any other model)
• Uses Ollama's native API format
• Streaming uses newline-delimited JSON
• Default URL: http://localhost:11434/api/chat

• No API key required
• Make sure LM Studio is running with the local server enabled (Developer tab)
• Load a model in LM Studio before using
• Uses OpenAI-compatible API format
• Streaming uses Server-Sent Events (SSE)
• Default URL: http://localhost:1234/v1/chat/completions
• Note: LM Studio can be slow, especially with large models. If you get timeout errors, increase the timeout:

  export LLM_TIMEOUT=600  # 10 minutes

The CLI provides clear error messages for common issues:

• Missing API key (for OpenRouter)
• Invalid provider
• Empty input
• Network errors (connection refused for local providers usually means the service isn't running)
• API errors
• Timeout errors: If you see "context deadline exceeded" or "request timed out", especially with LM Studio or large models, increase the timeout:

  export LLM_TIMEOUT=600  # Set to 600 seconds (10 minutes) or higher

• API keys are only read from environment variables (never from command-line arguments)
• API keys are never logged or stored
• Use environment variables for API keys in production environments
• Local providers (Ollama, LM Studio) don't require API keys and run entirely on your machine

Contributions are welcome! Please feel free to submit a Pull Request.

MIT License - see LICENSE file for details.

For issues and questions:

• Check the OpenRouter documentation
• Check the Ollama documentation
• Check the LM Studio documentation
• Open an issue on GitHub