AI Agents

MANYOYO supports multiple AI CLI tools (agents), providing shortcuts to launch YOLO/SOLO mode.

Note: Run profiles should be under runs.<name> in ~/.manyoyo/manyoyo.json; use absolute paths for envFile and map style for env.

Supported Agents

Claude Code

Anthropic's official Claude AI command-line tool.

Shortcuts:

bash

manyoyo run -y c          # Recommended
manyoyo run -y claude
manyoyo run -y cc

Equivalent to:

bash

manyoyo run -x claude --dangerously-skip-permissions

Resume session:

bash

manyoyo run -n <container-name> -- -c
manyoyo run -n <container-name> -- --continue

Configuration example:

json5

// runs.claude in ~/.manyoyo/manyoyo.json
{
    "envFile": ["/abs/path/anthropic_claudecode.env"],
    "yolo": "c"
}

Environment variables:

bash

# ~/.manyoyo/env/anthropic_claudecode.env
export ANTHROPIC_BASE_URL="https://api.anthropic.com"
export ANTHROPIC_AUTH_TOKEN="sk-xxxxxxxx"
export ANTHROPIC_MODEL="claude-sonnet-4-5"
export ANTHROPIC_DEFAULT_OPUS_MODEL="claude-opus-4-5"
export ANTHROPIC_DEFAULT_SONNET_MODEL="claude-sonnet-4-5"
export ANTHROPIC_DEFAULT_HAIKU_MODEL="claude-haiku-4-5"
export CLAUDE_CODE_SUBAGENT_MODEL="claude-sonnet-4-5"

Common commands:

bash

# Start YOLO mode
manyoyo run -r claude

# View version
manyoyo run -r claude -- --version

# View help
manyoyo run -r claude -- --help

# Resume last session
manyoyo run -r claude -- -c

Gemini

Google's Gemini AI command-line tool.

Shortcuts:

bash

manyoyo run -y gm         # Recommended
manyoyo run -y gemini
manyoyo run -y g

Equivalent to:

bash

manyoyo run -x gemini --yolo

Resume session:

bash

manyoyo run -n <container-name> -- -r
manyoyo run -n <container-name> -- --resume

Configuration example:

json5

// runs.gemini in ~/.manyoyo/manyoyo.json
{
    "envFile": ["/abs/path/gemini.env"],
    "yolo": "gm"
}

Environment variables:

bash

# ~/.manyoyo/env/gemini.env
export GEMINI_API_KEY="your-api-key"
export GEMINI_MODEL="gemini-2.0-flash-exp"

Common commands:

bash

# Start YOLO mode
manyoyo run -r gemini

# View version
manyoyo run -r gemini -- --version

# Resume session
manyoyo run -r gemini -- -r

Codex

OpenAI's Codex command-line tool.

Shortcuts:

bash

manyoyo run -y cx         # Recommended
manyoyo run -y codex

Equivalent to:

bash

manyoyo run -x codex --dangerously-bypass-approvals-and-sandbox

Resume session:

bash

manyoyo run -n <container-name> -- resume --last
manyoyo run -n <container-name> -- resume <session-id>

Configuration example:

json5

// runs.codex in ~/.manyoyo/manyoyo.json
{
    "envFile": ["/abs/path/openai_codex.env"],
    "volumes": [
        "/Users/pc_user/.codex/auth.json:/root/.codex/auth.json"
    ],
    "yolo": "cx"
}

Environment variables:

bash

# ~/.manyoyo/env/openai_codex.env
export OPENAI_BASE_URL=https://chatgpt.com/backend-api/codex

Common commands:

bash

# Start YOLO mode
manyoyo run -r codex

# View session list
manyoyo run -r codex -- list

# Resume last session
manyoyo run -r codex -- resume --last

# Resume specific session
manyoyo run -r codex -- resume <session-id>

OpenCode

Open-source AI code assistant.

Shortcuts:

bash

manyoyo run -y oc         # Recommended
manyoyo run -y opencode

Equivalent to:

bash

manyoyo run -x "OPENCODE_PERMISSION='{\"*\":\"allow\"}' opencode"

Resume session:

bash

manyoyo run -n <container-name> -- -c
manyoyo run -n <container-name> -- --continue

Configuration example:

json5

// runs.opencode in ~/.manyoyo/manyoyo.json
{
    "envFile": ["/abs/path/opencode.env"],
    "yolo": "oc"
}

Environment variables:

bash

# ~/.manyoyo/env/opencode.env
export OPENAI_API_KEY="your-api-key"
export OPENAI_BASE_URL="https://api.openai.com/v1"

Common commands:

bash

# Start YOLO mode
manyoyo run -r opencode

# View version
manyoyo run -r opencode -- --version

# Resume session
manyoyo run -r opencode -- -c

YOLO Mode Explanation

YOLO (You Only Live Once) mode refers to AI agents skipping permission confirmation and automatically executing commands.

Why Use YOLO Mode?

Advantages:

Improves efficiency, reduces interaction
Suitable for automation scenarios
Runs in isolated containers, protecting host machine security

Risks:

AI may execute dangerous commands (e.g., rm -rf)
In MANYOYO containers, risks are limited to inside the container

Security Isolation

MANYOYO provides secure container isolation:

Host Machine
  └─ MANYOYO Container (Isolated environment)
      └─ AI Agent (YOLO mode)
          ├─ File operations → Only affects container
          ├─ Process operations → Only affects container
          └─ Network operations → Configurable isolation

Protection mechanisms:

Container filesystem isolation
Resource limits
Network isolation (optional)
Can delete and restart containers at any time

Agent Comparison

Agent	Shortcut	Resume Command	Primary Use	Supported Languages
Claude Code	`-y c`	`-- -c`	General programming assistance	Multi-language
Gemini	`-y gm`	`-- -r`	General programming assistance	Multi-language
Codex	`-y cx`	`-- resume --last`	Code generation	Multi-language
OpenCode	`-y oc`	`-- -c`	Open-source code assistant	Multi-language

Session Management

Create New Session

bash

# Create new session (auto-generate container name)
manyoyo run -y c

# Create named session
manyoyo run -n my-session -y c

Resume Session

Different agents have different resume methods:

bash

# Claude Code
manyoyo run -n my-session -- -c

# Gemini
manyoyo run -n my-session -- -r

# Codex
manyoyo run -n my-session -- resume --last

# OpenCode
manyoyo run -n my-session -- -c

Session Persistence

Container state determines whether sessions are preserved:

bash

# Exit and keep container running (session preserved)
# Select 'y' in interactive prompt

# Remove container (session lost)
manyoyo rm my-session

View Sessions

bash

# List all container sessions
manyoyo ps

# View specific container
docker ps -a | grep my-session

Switching Between Agents

Switch Within Container

bash

# Start Claude Code
manyoyo run -n dev -y c

# After exit, enter shell
manyoyo run -n dev -x /bin/bash

# Manually run other agents in shell
gemini --yolo
codex --dangerously-bypass-approvals-and-sandbox

Use Different Containers

bash

# Claude Code container
manyoyo run -n claude-session -y c

# Codex container
manyoyo run -n codex-session -y cx

# Switch as needed
manyoyo run -n claude-session -- -c
manyoyo run -n codex-session -- resume --last

Cycling Between Agent and /bin/bash

MANYOYO supports flexible switching between AI agents and shell:

Switch from Agent to Shell

bash

# Start agent
manyoyo run -n dev -y c

# After working, exit agent

# Select 'i' to enter interactive shell
# Or use command
manyoyo run -n dev -x /bin/bash

Switch from Shell to Agent

bash

# In shell
manyoyo run -n dev -x /bin/bash

# Run agent directly inside container
claude --dangerously-skip-permissions
gemini --yolo
codex --dangerously-bypass-approvals-and-sandbox

# Or exit and use command
manyoyo run -n dev -y c

Workflow Example

bash

# 1. Start Claude Code for development
manyoyo run -n project -y c

# 2. AI helps write code...

# 3. Exit, enter shell to check
manyoyo run -n project -x /bin/bash

# 4. Manually test in shell
$ npm test
$ git status
$ ls -la

# 5. Continue using AI
$ claude --dangerously-skip-permissions

# 6. Or exit and resume
manyoyo run -n project -- -c

Tips and Best Practices

Use Run Configurations

Create dedicated configuration for each agent:

bash

# Create configuration
cat > ~/.manyoyo/manyoyo.json << 'EOF'
{
    "runs": {
        "claude": {
            "envFile": ["/abs/path/anthropic_claudecode.env"],
            "yolo": "c"
        }
    }
}
EOF

# Use configuration (simple)
manyoyo run -r claude

Unified Container Naming

Use meaningful container names:

bash

# Name by project
manyoyo run -n webapp-claude -r claude
manyoyo run -n api-codex -r codex

# Name by function
manyoyo run -n dev-claude -r claude
manyoyo run -n test-gemini -r gemini

Multi-Agent Collaboration

Use multiple agents in the same project:

bash

# Claude for architecture design
manyoyo run -n project-claude --hp ~/project -r claude

# Codex for code generation
manyoyo run -n project-codex --hp ~/project -r codex

# Switch usage
manyoyo run -n project-claude -- -c
manyoyo run -n project-codex -- resume --last

Configure Environment Isolation

Configure different environments for different agents:

bash

# Development environment - Use Claude
cat > ~/.manyoyo/manyoyo.json << 'EOF'
{
    "runs": {
        "dev": {
            "envFile": ["/abs/path/anthropic_dev.env"],
            "env": {
                "NODE_ENV": "development"
            },
            "yolo": "c"
        },
        "prod": {
            "envFile": ["/abs/path/gemini_prod.env"],
            "env": {
                "NODE_ENV": "production"
            },
            "yolo": "gm"
        }
    }
}
EOF

Troubleshooting

Agent Cannot Start

Check environment variables:

bash

# Verify environment variables
manyoyo config show -r claude

# Test environment variables
manyoyo run -r claude -x 'env | grep ANTHROPIC'

Check image:

bash

# Confirm agent is installed in image
manyoyo run -x which claude
manyoyo run -x which gemini
manyoyo run -x which codex

Session Cannot Resume

Check container status:

bash

# Check if container exists
manyoyo ps
docker ps -a | grep <container-name>

# View container logs
docker logs <container-name>

Use correct resume command:

bash

# Claude Code: -c or --continue
manyoyo run -n test -- -c

# Gemini: -r or --resume
manyoyo run -n test -- -r

# Codex: resume --last
manyoyo run -n test -- resume --last

API Authentication Failed

Check API Key:

bash

# View environment file
cat ~/.manyoyo/env/anthropic_claudecode.env

# Test API
curl -H "x-api-key: $ANTHROPIC_AUTH_TOKEN" \
     https://api.anthropic.com/v1/messages

Update configuration:

bash

# Edit environment file
vim ~/.manyoyo/env/anthropic_claudecode.env

# Restart container
manyoyo rm test
manyoyo run -n test -r claude

Basic Usage - Learn basic commands and operations
Configuration Examples - View agent configuration examples
Environment Variables - Learn how to configure environment variables
Runtime Issues - AI CLI tool troubleshooting

AI Agents ​

Supported Agents ​

Claude Code ​

Gemini ​

Codex ​

OpenCode ​

YOLO Mode Explanation ​

Why Use YOLO Mode? ​

Security Isolation ​

Agent Comparison ​

Session Management ​

Create New Session ​

Resume Session ​

Session Persistence ​

View Sessions ​

Switching Between Agents ​

Switch Within Container ​

Use Different Containers ​

Cycling Between Agent and /bin/bash ​

Switch from Agent to Shell ​

Switch from Shell to Agent ​

Workflow Example ​

Tips and Best Practices ​

Use Run Configurations ​

Unified Container Naming ​

Multi-Agent Collaboration ​

Configure Environment Isolation ​

Troubleshooting ​

Agent Cannot Start ​

Session Cannot Resume ​

API Authentication Failed ​

Related Documentation ​

AI Agents

Supported Agents

Claude Code

Gemini

Codex

OpenCode

YOLO Mode Explanation

Why Use YOLO Mode?

Security Isolation

Agent Comparison

Session Management

Create New Session

Resume Session

Session Persistence

View Sessions

Switching Between Agents

Switch Within Container

Use Different Containers

Cycling Between Agent and /bin/bash

Switch from Agent to Shell

Switch from Shell to Agent

Workflow Example

Tips and Best Practices

Use Run Configurations

Unified Container Naming

Multi-Agent Collaboration

Configure Environment Isolation

Troubleshooting

Agent Cannot Start

Session Cannot Resume

API Authentication Failed

Related Documentation