Installing Arcade Locally
This guide will help you install Arcade and set up your development environment. If you are developing tools to use with Arcade, this guide will provide you with a complete local deployment of Arcade for developing and testing tools.
Prerequisites
Before you begin, make sure you have the following:
- Python 3.10 or higher
- pip: The Python package installer should be available. It’s typically included with Python.
- Arcade Account: Sign up for an Arcade account if you haven’t already.
- Package Manager: Either Brew (macOS) or Apt (linux) to install the engine binary.
Verify your Python version by running python --version or python3 --version in your terminal.
Installation
Install the Client
To connect to the cloud or local Arcade Engine, we need to install the Arcade CLI from the arcade-ai package.
pip install arcade-ai
arcade loginFor a simple example on using the Arcade CLI, see the quickstart on building tools
Install the Engine
To run the Arcade Engine locally, you’ll need to install arcade-engine. Choose the installation method that matches your operating system.
This will install a template engine configuration.
macOS (Homebrew)
brew install ArcadeAI/tap/arcade-engineInstall a MCP Server
In order to run the Arcade worker, you’ll need to install at least one tool. For local development, you can just pip install a tool in the same environment as the client.
pip install arcade-mathFor more information on installing tools, see the Toolkit Installation page.
To see all available tools, view the MCP Servers Page.
Set OpenAI API key
Before starting the Engine, we need to set an OpenAI API Key that the Engine can use to connect to OpenAI.
Edit the engine.env file to set the OPENAI_API_KEY environment variable:
OPENAI_API_KEY="<your_openai_api_key>"See the configuration overview for more information on how to configure Arcade Engine and how to locate the engine.env file.
Start the Engine and worker
To run both the engine and a local worker, follow these steps:
- First, start the worker:
arcade serve- Then, start the engine:
arcade-engineThe Engine and worker should both be running locally now.
To run the Engine on it’s own, you can run:
arcade-engineNote that the Engine requires at least one Arcade worker to run.
Connect
To chat with the running Engine, in a separate terminal instance, run:
arcade chat -h localhostYou are now chatting with Arcade locally!
Next Steps
- Building Tools: Learn how to build tools with your local Arcade Instance in the Creating a Toolkit guide.
- Hosting With Docker: Learn how to run the Engine in Docker.
Troubleshooting
Engine Binary Not Found
❌ Engine binary not foundor
command not found: arcade-engineThis means that the Arcade Engine cannot be found in your path. Brew and Apt will automatically add the binary to you path.
Check that the binary has been properly installed (These are the common installation locations):
Brew
ls $HOMEBREW_REPOSITORY/Cellar/arcade-engine/<version>/bin/arcade-engineApt
ls /usr/bin/arcade-engineIf the binary is found, add it to your path with:
export PATH=$PATH:/path/to/your/binaryMCP Servers Not Found
No toolkits found in Python environment. Exiting...This means that there are no MCP Servers found in the same environment as the Arcade SDK. Ensure that you are installing the MCP Server package in the same environment and see the MCP Server Installation Guide for more details.
Engine Config Not Found
❌ Config file 'engine.yaml' not found in any of the default locations.or
Arcade Engine has finished with error: unable to read config file $HOME/.arcade/engine.yaml: open $HOME/.arcade/engine.yaml: no such file or directoryThe engine config may be located in one of the following directories:
- $HOME/.arcade/engine.yaml
- $HOMEBREW_REPOSITORY/etc/arcade-engine/engine.yaml (Homebrew)
- /etc/arcade-ai/engine.yaml (Apt)
The engine config will be downloaded by and added to one of these locations when installing the engine.
When running the engine, the config needs to be in the $HOME/.arcade/ directory or explicitly located with arcade-engine -c /path/to/engine.yaml.
If you cannot find your engine config, you can save and use the Configuration Templates.