Quick Start

Get Agent Swarm x Code Evolution running in 5 minutes. Viben CLI enables AI agent swarms to continuously evolve and optimize your codebase through FileEvo, intelligent task management, and collaborative agent orchestration.

Step 1: Install Viben CLI

npm install -g @viben/cli

Verify the installation:

viben --help

Step 2: Initialize Your First Workspace

Navigate to your project directory and initialize Viben:

cd /path/to/your/project
viben init

Output:

Initialized Viben workspace in /path/to/your/project
  Created .viben/config.yaml

Next steps:
  viben provider create -t anthropic    # Set up API provider
  viben mcp install <name>              # Install MCP server
  viben skill install <name>            # Install skill

Step 3: Configure API Provider

Set up your preferred AI provider. Viben supports multiple providers including Anthropic, OpenAI, Google, Azure, and more.

Option A: Using Environment Variables (Recommended)

Set API keys as environment variables:

# Anthropic
export ANTHROPIC_API_KEY="sk-ant-xxx"

# OpenAI
export OPENAI_API_KEY="sk-xxx"

Then create the provider:

viben provider create -t anthropic

Option B: Provide API Key Directly

viben provider create -t anthropic --api-key "sk-ant-xxx"

提示

When providing an API key directly, it will be encrypted and securely stored in ~/.viben/models.yaml.

Verify Provider

Check provider connectivity:

viben provider status

Output:

Provider Status:
  anthropic-main   anthropic   ✓ connected   latency: 120ms

Step 4: Create Your First Agent

Create an AI agent instance:

viben agent create -n my-agent

Output:

Agent: my-agent
Type: claude-code
Created: 2024-01-15

Paths:
  Config:   ~/.viben/agents/my-agent/config.yaml
  Memory:   ~/.viben/agents/my-agent/memory/
  Sessions: ~/.viben/agents/my-agent/.agent_sessions/

Configure Agent

Set a model for your agent:

viben agent config -n my-agent --set model=claude-sonnet-4-20250514

Set as Default

Set this agent as the default:

viben agent set-default -n my-agent

Step 5: Install MCP Servers

Install commonly used MCP servers:

# Filesystem access
viben mcp add filesystem --agent my-agent --command npx --args @anthropic-ai/mcp-server-filesystem /home/user

# Git operations
viben mcp add git --agent my-agent --command npx --args @anthropic-ai/mcp-server-git

List installed MCP servers:

viben mcp list --agent my-agent

Output:

MCP Servers for Agent: my-agent
  Name         Command                              Enabled
  filesystem   npx @anthropic-ai/mcp-server-fs      yes
  git          npx @anthropic-ai/mcp-server-git     yes

Step 6: Verify Setup

Check overall status:

viben agent status

Output:

Agent: my-agent (default)
Type: claude-code
Model: claude-sonnet-4-20250514 (anthropic-main)

MCP: filesystem, git (2 enabled)
Skills: none

Memory:
  MEMORY.md     0 KB    empty

Sessions: 0

Common Workflows

View All Configuration

# List all configuration (show global and workspace)
viben config list --show-origin

Edit Configuration

# Open configuration in editor
viben config edit

# Or set specific values
viben config set settings.editor vim

Manage Multiple Agents

# List all agents
viben agent list

# Create agent from template
viben agent create -n research-bot -f coding-assistant

# Switch default agent
viben agent set-default -n research-bot

Configure Model Aliases

Set up convenient model aliases:

# Create aliases for quick reference
viben model alias create -n fast -m claude-3-5-haiku-latest
viben model alias create -n smart -m claude-sonnet-4-20250514
viben model alias create -n best -m claude-opus-4-20250514

Now you can use fast, smart, or best instead of full model names:

viben agent config -n my-agent --set model=smart

Set Up Model Fallback Chain

Configure backup models in case the primary model is unavailable:

viben model set-default -n your-preferred-model

Workspace-Specific Configuration

Override global settings for a specific project:

# In your project directory
viben config set --workspace mcp.enabled '["filesystem", "git", "browser"]'

JSON Output for Automation

All commands support the --json flag for scripting and AI agent integration:

# Get agent list as JSON
viben agent list --json

# Get provider status as JSON
viben provider status --json

Example JSON output:

{
  "success": true,
  "data": {
    "current": "my-agent",
    "agents": [
      {
        "id": "my-agent",
        "name": "My Agent",
        "type": "claude-code",
        "path": "~/.viben/agents/my-agent/"
      }
    ]
  }
}

Quick Reference

Task	Command
Initialize workspace	viben init
Create provider	viben provider create -t <type>
Create agent	viben agent create -n <name>
Configure agent	viben agent config -n <name> --set <key>=<value>
Add MCP	viben mcp add <name> --agent <id> --command <cmd>
Install skill	viben skill install <name>
Check status	viben agent status
List agents	viben agent list
List providers	viben provider list
List models	viben model list
Start Gateway	viben gateway start
View executors	viben executor list

Next Steps

Now that you have Viben CLI set up, you can:

• Explore Agent Management for the complete agent lifecycle
• Configure additional API Providers for more model options
• Install more MCP Servers to extend functionality
• Learn about Gateway runtime architecture