@aigne/cli
Command-line tool for AIGNE Framework, providing convenient development and management capabilities.
Introduction#
@aigne/cli is the official command-line tool for AIGNE Framework, designed to simplify the development, testing, and deployment processes for AIGNE applications. It provides a series of useful commands to help developers quickly create projects, run agents, test code, and deploy applications.
Features#
- Project Creation: Quickly create new AIGNE projects with predefined file structures and configurations
- Agent Running: Easily run and test AIGNE agents
- Testing Support: Built-in test command for unit testing and integration testing
- MCP Services: Support for launching agents as MCP servers for integration with external systems
- Interactive Interface: Beautiful command-line interface providing an intuitive user experience
- Multi-model Support: Support for OpenAI, Claude, XAI, and other model providers
Installation#
Using npm#
npm install -g @aigne/cli
Using yarn#
yarn global add @aigne/cli
Using pnpm#
pnpm add -g @aigne/cli
Basic Commands#
AIGNE CLI provides the following main commands:
# Display help information
aigne --help
# Create a new project
aigne create [path]
# Run an agent
aigne run --path xxx
# Run tests
aigne test --path xxx
# Start MCP server
aigne serve-mcp --path xxx
# Start observability server
aigne observe [option]
Create Command#
Create a new AIGNE project with agent configuration files.
# Create a project in the current directory (will prompt for project name)
aigne create
# Create a project at the specified path
aigne create my-project
The interactive creation process will ask for:
- Project name
- Project template (currently supports the default template)
Run Command#
Launch a chat loop with the specified agent.
# Run the agent in the current directory
aigne run
# Run the agent at the specified path
aigne run path/to/agents
# Run the agent from a remote URL
aigne run https://example.com/aigne-project
# Run a specific agent
aigne run --entry-agent myAgent
Available options:
--entry-agent <entry-agent>- Specify the agent name to run (defaults to the first agent found)--cache-dir <dir>- Specify the directory to download the package to (used in URL mode)--model <provider[:model]>- Specify the AI model in format 'provider[:model]' where model is optional (e.g., 'openai' or 'openai:gpt-4o-mini')--verbose- Enable verbose logging
Test Command#
Run tests in the specified agents directory.
# Test the agents in the current directory
aigne test
# Test the agents at the specified path
aigne test path/to/agents
Serve MCP Command#
Serve the agents in the specified directory as a MCP server.
# Start MCP server on default port 3000
aigne serve-mcp
# Start MCP server on specified port
aigne serve-mcp --port 3001
# Start MCP server for agents at specified path
aigne serve-mcp --path path/to/agents
Serve Command (observability)#
Start the service for monitoring data
# Start observability server on default port 7890
aigne observe
# Start observability server on specified port
aigne observe --port 3001
License#
Elastic-2.0