DeepAgents CLI

A terminal-based AI coding assistant with built-in tools for file operations, shell commands, web search, and subagent delegation. Features customizable skills, persistent memory, and automatic project detection.

What This Skill Does

This skill helps you install, configure, and use the DeepAgents CLI - a powerful terminal-based AI coding assistant that provides:

Instructions

Step 1: Check Installation

First, verify if the DeepAgents CLI is already installed:

```bash

which deepagents

```

If not found, proceed to installation.

Step 2: Install DeepAgents CLI

Choose one of the following installation methods based on the user's package manager:

```bash

Using npm

npm install -g deepagents-cli

Using pnpm

pnpm add -g deepagents-cli

Using yarn

yarn global add deepagents-cli

```

For one-time use without installation, users can run:

```bash

npx deepagents-cli

```

Step 3: Configure API Keys

DeepAgents CLI requires at least one model provider API key. Help the user set up their environment variables:

**Required (at least one):**

```bash

export ANTHROPIC_API_KEY="your-anthropic-key"

export OPENAI_API_KEY="your-openai-key"

export GOOGLE_API_KEY="your-google-key"

```

**Optional (for enhanced features):**

```bash

Web search capability

export TAVILY_API_KEY="your-tavily-key"

LangSmith tracing for debugging

export LANGCHAIN_TRACING_V2=true

export LANGCHAIN_API_KEY="your-langsmith-key"

```

Suggest adding these to the user's shell profile (~/.bashrc, ~/.zshrc, etc.) for persistence.

Step 4: Basic Usage

Guide the user through basic DeepAgents CLI usage:

**Start the interactive CLI:**

```bash

deepagents

```

**Get help:**

```bash

deepagents help

```

**List available agents:**

```bash

deepagents list

```

Step 5: Advanced Usage Options

Explain advanced features based on user needs:

**Use a specific agent configuration:**

```bash

deepagents --agent mybot

```

**Use a specific model:**

```bash

deepagents --model claude-sonnet-4-5-20250929

deepagents --model gpt-4o

```

**Auto-approve tool usage (skip confirmation prompts):**

```bash

deepagents --auto-approve

```

**Execute code in a remote sandbox:**

```bash

deepagents --sandbox modal # Options: modal, runloop, daytona

deepagents --sandbox-id dbx_123 # Reuse existing sandbox

```

Step 6: Skills Management

The DeepAgents CLI supports a customizable skill system. Show users how to manage skills:

**List all skills (global + project-specific):**

```bash

deepagents skills list

```

**Create a new skill:**

```bash

deepagents skills create my-skill

```

**View skill information:**

```bash

deepagents skills info web-research

```

Step 7: Troubleshooting

If users encounter issues:

**Binary not found error:**

```bash

Reinstall the package

npm uninstall -g deepagents-cli

npm install -g deepagents-cli

```

**Unsupported platform:**

Check if the user's platform is supported:

If unsupported, direct them to open an issue at https://github.com/langchain-ai/deepagentsjs/issues

Usage Examples

Example 1: Basic Interactive Session

```bash

Start DeepAgents with Anthropic Claude

export ANTHROPIC_API_KEY="sk-ant-..."

deepagents

```

Example 2: Quick Task with Specific Model

```bash

Use GPT-4o for a specific task

deepagents --model gpt-4o --auto-approve

```

Example 3: Project-Specific Agent with Skills

```bash

Navigate to project directory (auto-detects project config)

cd /path/to/my-project

deepagents --agent project-bot

List available skills for this project

deepagents skills list

```

Example 4: Sandboxed Execution

```bash

Execute code in a Modal sandbox

deepagents --sandbox modal --auto-approve

```

Important Notes

Related Resources