AgentKit CLI Overview

AgentKit CLI is a powerful command-line tool designed for developers to simplify and accelerate the full lifecycle of Agent application development, deployment, and management. Whether you’re rapidly prototyping locally or deploying at scale in production, the agentkit command delivers a consistent and efficient developer experience.

Key Advantages

• Declarative configuration: Manage all configuration through a single agentkit.yaml file—clear, portable, and easy to version-control.
• Multi-environment deployment: Seamlessly supports three modes—local development, hybrid deployment, and fully cloud deployment—to fit different scenarios.
• One-command workflows: High-level commands like agentkit launch combine multiple steps (build, deploy, etc.) into one, dramatically improving efficiency.
• Templated projects: Built-in application templates help you bootstrap quickly without starting from scratch.
• Smart interactivity: Interactive wizards such as agentkit config guide you through setup and reduce the learning curve.

Main Commands

The agentkit CLI provides a set of intuitive commands to manage your Agent applications:

Core Commands

Command	Description
agentkit init	Initialize a new Agent project, with optional templates.
agentkit config	Configure agentkit.yaml via an interactive wizard.
agentkit build	Build your Agent application into a Docker image.
agentkit deploy	Deploy a built image to the specified environment (local or cloud).
agentkit launch	Build and deploy in one command—the most commonly used command during development.
agentkit invoke	Interact with a deployed Agent for quick testing.
agentkit status	View the Agent runtime status and endpoint.
agentkit destroy	Stop and destroy the Agent instance, releasing related resources.

Platform Service Commands

Command	Description
agentkit memory	Manage AgentKit Memory collections.
agentkit knowledge	Manage AgentKit Knowledge bases.
agentkit tools	Manage AgentKit Tools and Sessions.
agentkit runtime	Manage AgentKit Runtime instances.

Want detailed usage for each command? See Command Reference.

Three Deployment Modes

One of the core features of AgentKit CLI is its flexible deployment model, designed to adapt to different stages from development to production.

1. Local Mode (local)

Build and deploy on your local machine—ideal for fast development and debugging.

• Workflow: Source code → Local Docker build → Run local container
• Benefits: Fast iteration, supports offline work, easy to debug directly.

2. Cloud Mode (cloud)

Fully host build and deployment on the AgentKit Platform—a best practice for production.

• Workflow: Source code → Cloud auto build → Cloud image registry → Production-grade Runtime
• Benefits: No local Docker required, built-in observability, high environment consistency, secure and reliable.

3. Hybrid Mode (hybrid)

Build the image locally, then push it to the cloud AgentKit Platform for deployment. This combines local control with cloud capabilities.

• Workflow: Source code → Local Docker build → Upload to cloud image registry → Production-grade Runtime
• Benefits: Balances local build flexibility with the stability of a cloud runtime—suitable for scenarios that require a custom build process.

Configuration File (agentkit.yaml)

agentkit.yaml is the core of an AgentKit project. It defines all application configuration in a declarative manner. We strongly recommend managing this file with the agentkit config command.

# General configuration
common:
  agent_name: my-agent
  entry_point: app.py
  language_version: '3.12'
  launch_type: cloud # Define the default deployment mode

# Mode-specific configuration
launch_types:
  cloud:
    region: cn-beijing
    image_tag: latest
    runtime_envs:
      MODEL_API_KEY: "your-api-key"

Want detailed explanations of all configuration options? See Configuration Reference.

Quick Start

In just a few minutes, you can deploy your first Agent.

# 1. Initialize a project (choose the Basic App template)
mkdir my-first-agent && cd my-first-agent
agentkit init my-first-agent

# Or specify a particular template
agentkit init my-first-agent --template basic_stream  # Streaming output template
agentkit init my-first-agent --template a2a           # A2A protocol template

# 2. Configure the app (via the interactive wizard)
agentkit config

# 3. Deploy to the cloud in one command
agentkit launch

# 4. Test your Agent
agentkit invoke "Hello!"

# 5. Check runtime status
agentkit status

# 6. Clean up resources (optional)
agentkit destroy

Explore More Features

# List available templates
agentkit init --list-templates

# Manage platform services
agentkit memory list      # List memory collections
agentkit knowledge list   # List knowledge bases
agentkit tools list       # List tools
agentkit runtime list     # List runtime instances

Environment Requirements

✅ Basic Requirements (All Modes)

• Python 3.10 or later
• pip package manager
• AgentKit SDK installed

🐍 Python Development Environment

• Supports Python 3.10, 3.11, 3.12, 3.13
• Recommended to manage dependencies with requirements.txt

🐹 Golang Development Environment

• Supports Golang 1.24
• Manage dependencies with go.mod
• Suitable for high-performance scenarios

🐳 Local Development (Local/Hybrid Mode)

• Docker Desktop or Docker Engine 20.10+
• Ensure the Docker service is running

☁️ Cloud Deployment (Cloud/Hybrid Mode)

• A Volcengine account
• Configure access credentials (AK/SK). We recommend using the secure global configuration approach:

  # Initialize global configuration
  agentkit config --global --init
  
  # Set access credentials
  agentkit config --global --set volcengine.access_key="your_access_key"
  agentkit config --global --set volcengine.secret_key="your_secret_key"
  
  # (Optional for Byteplus) Set BytePlus as the default cloud provider
  agentkit config --global --set defaults.cloud_provider=byteplus
  
  # (Optional for Byteplus) Set Byteplus access credentials
  agentkit config --global --set byteplus.access_key="your_byteplus_access_key"
  agentkit config --global --set byteplus.secret_key="your_byteplus_secret_key"

Next Steps

• 📖 Command Reference: Dive into parameters and usage for each CLI command.
• ⚙️ Configuration Reference: Master all configuration options in agentkit.yaml.
• 🚀 Quick Start: Follow an end-to-end tutorial to build a complete Agent application from scratch.