cade-cli 0.8.0

Cade - The CLI Agent from Arcade.dev

Cade

An AI-powered CLI agent for coding and everyday tasks. Powered by Arcade.dev.

Installation

Prerequisites

• Python 3.11+
• Arcade account (arcade.dev)
• AI provider API key: OPENAI_API_KEY or ANTHROPIC_API_KEY

Homebrew (macOS/Linux)

brew tap ArcadeAI/tap
brew install cade

Install with uv

curl -LsSf https://astral.sh/uv/install.sh | sh
uv tool install cade-cli

Install with pip

pip install cade-cli

From Source

git clone https://github.com/arcadeai-labs/cade.git
cd cade
uv venv --python 3.11
source .venv/bin/activate
uv sync

Authenticate

cade login

Usage

Start a Chat

cade

Resume a Thread

cade -r                        # Resume most recent
cade resume "my-project"       # Resume by name

Authentication

Cade uses Arcade Cloud for authentication and shares credentials with arcade-cli.

cade login                      # Log in to Arcade Cloud
cade logout                     # Log out
cade whoami                     # Show current login status

Context Management

Switch between organizations and projects for Arcade Cloud features.

cade context show               # Show current org/project
cade context list               # List available orgs and projects
cade context switch -i          # Interactive selection
cade context switch --org my-org --project my-project

Single Message Mode

cade -m "What files are in this directory?"
cat error.log | cade -m "What went wrong?"

Options

Option	Description
-r, --resume	Resume the most recent thread
-m, --message	Single message mode (non-interactive)
-L, --local-only	Disable remote tools (use only local tools)
-v, --verbose	Enable debug logging
--version	Show version

In-Chat Commands

Command	Description
/help	Show available commands
/logs	View recent log entries
/clear	Clear the screen
/copy	Copy last response to clipboard
Ctrl+C	Exit

Thread Management

cade thread list                          # List all threads
cade thread list --branch main            # Filter by branch
cade thread get <thread-id>               # Get thread details
cade thread get <thread-id> --messages    # Show messages
cade thread delete <thread-id>            # Delete thread

Tool Management

Tools come from three sources: local, Arcade Cloud, and MCP servers.

cade tool list                    # List all tools
cade tool list --source local     # Filter by source
cade tool search "file"           # Search tools
cade tool info Local_ReadFile     # Tool details

Built-in Tools

Tool	Description
Local_ReadFile	Read file contents
Local_WriteFile	Write or append to files
Local_ListFiles	List directory contents
Local_SearchText	Search for text patterns
Local_ExecuteShell	Run shell commands
Local_CreateDirectory	Create directories
Local_DeleteFile	Delete files
Local_GetGitStatus	Get git status

MCP Servers

Connect to MCP servers for extended capabilities.

cade mcp list                                                    # List servers
cade mcp add my-server http://localhost:8080                     # Add server
cade mcp add my-server http://localhost:8080 --auth bearer -t <token>  # With auth
cade mcp test my-server                                          # Test connection
cade mcp enable my-server                                        # Enable
cade mcp disable my-server                                       # Disable
cade mcp rm my-server                                            # Remove

Configuration

Config is stored in ~/.cadecoder/:

File	Description
cadecoder.toml	Settings
cadecoder_history.db	Thread history
cadecoder.log	Logs
mcp_servers.yaml	MCP server configs

Environment Variables

Variable	Description
OPENAI_API_KEY	OpenAI API key
OPENAI_BASE_URL	Custom OpenAI-compatible API endpoint
ANTHROPIC_API_KEY	Anthropic API key
ARCADE_API_KEY	Arcade API key (alternative to OAuth)
ARCADE_BASE_URL	Custom Arcade API endpoint
CADE_LOCAL_ONLY	Set to 1 to disable remote tools
CADECODER_HOME	Override config directory

Example Config

# ~/.cadecoder/cadecoder.toml

default_model = "gpt-4.1"
debug_mode = false
use_responses_api = true

[responses_config]
enabled = true
streaming_enabled = true

[model_settings]
provider = "openai"
model = "gpt-4.1"

[tool_settings]
# Tool filtering is managed via MCP server configuration
# See: cade mcp add --help

Using Local or Custom LLMs

Cade works with any OpenAI-compatible API, including local servers (Ollama, vLLM, llama.cpp) and alternative cloud providers (Together AI, Groq, Fireworks).

Local-Only Mode

When using local LLMs, you can skip Arcade Cloud authentication entirely with --local-only:

# Local Ollama server without Arcade Cloud
cade chat --local-only --endpoint "http://localhost:11434/v1" --model "llama3"

# Or via environment variable
CADE_LOCAL_ONLY=1 cade chat --endpoint "http://localhost:11434/v1" --model "llama3"

This disables remote tools and uses only local tools. Cade will also gracefully fall back to local-only mode if Arcade Cloud authentication is not configured.

Via CLI Flags

# Local Ollama server
cade chat --endpoint "http://localhost:11434/v1" --model "glm-4.7-flash:latest"

# vLLM server
cade chat -e http://localhost:8000/v1 -m mistral-7b

Via Environment Variables

export OPENAI_BASE_URL="http://localhost:11434/v1"
export OPENAI_API_KEY="ollama"  # Dummy key for local model
cade chat --model glm-4.7-flash:latest

Via Config File

# ~/.cadecoder/cadecoder.toml

default_model = "glm-4.7-flash:latest"

[model_settings]
host = "http://localhost:11434/v1"
api_key = "ollama"

After configuring the TOML file:

cade chat

cade chat Configuration Precedence

Settings are resolved in this order (first is used):

1. CLI flags (--endpoint, --model)
2. Environment variables (OPENAI_BASE_URL, OPENAI_API_KEY)
3. Config file (model_settings.host, model_settings.api_key)
4. Hardcoded defaults

Contributing

Development Setup

git clone https://github.com/arcadeai-labs/cade.git
cd cade
uv sync --extra dev

Run Tests

pytest
ruff check src/
ruff format src/

Code Style

• Python 3.11+ with modern type hints (dict, list, | None)
• Ruff for linting and formatting
• Pytest for testing
• Docstrings for public functions and classes

Submitting Changes

1. Fork the repository
2. Create a feature branch
3. Make changes with tests
4. Run ruff check . && pytest
5. Open a Pull Request

Resources

• arcade.dev
• Documentation
• Issues
• Releases

License

MIT