Install a tool or MCP Server

Running a Local worker

To run a local worker without the arcade engine, you can run the following command locally:


arcade serve

This will check for any MCP Servers installed in the current python virtual environment and register them with the worker.

You can also pass in the --host and --port flags to specify the host and port to run the worker on. The default host is 127.0.0.1 and the default port is 8002.

PyPI Installation

All Arcade MCP Servers are available on PyPI. To install a MCP Server, run the following in the same python virtual environment as the arcade-ai package:


pip install arcade-[mcp_server_name]

For example, to install the math MCP Server, run:


pip install arcade-math

To verify the installation, run:


arcade show --local

which should return output similar to:


┏━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━┳━━━━━━━━━┓
┃ Name         ┃ Description                  ┃ Toolkit ┃ Version ┃
┡━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━╇━━━━━━━━━┩
│ Math.Add     │ Add two numbers together     │  Math   │  0.1.0  │
└──────────────┴──────────────────────────────┴─────────┴─────────┘

These are all tools that are installed in the same python virtual environment as the arcade-ai package.

See our MCP Servers Overview page for all available MCP Servers and individual installation instructions.

Local Package Installation

Locally built MCP Servers can also be installed with:


pip install .


pip install <wheel_name>

in the same python virtual environment as the arcade-ai package.

Hosted MCP Servers

To add a MCP Server to a hosted worker such as FastAPI, you can register them in the worker itself. This allows you to explicitly define which tools should be included on a particular worker.


import arcade_math
from fastapi import FastAPI
from arcade_tdk import Toolkit
from arcade_serve.fastapi import FastAPIWorker
 
app = FastAPI()
 
worker_secret = os.environ.get("ARCADE_WORKER_SECRET")
worker = FastAPIWorker(app, secret=worker_secret)
 
worker.register_toolkit(Toolkit.from_module(arcade_math))

Showing Tools From a Hosted Engine

To show all tools that are available on an engine, you can run


arcade show -h <engine_host> -p <engine_port>


┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━┳━━━━━━━━━┓
┃ Name                                   ┃ Description                                                               ┃ Package  ┃ Version ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━╇━━━━━━━━━┩
│ Github.CountStargazers                 │ Count the number of stargazers (stars) for a GitHub repository.           │ Github   │ 0.0.15  │
│ Github.CreateIssue                     │ Create an issue in a GitHub repository.                                   │ Github   │ 0.0.15  │
│ Github.CreateIssueComment              │ Create a comment on an issue in a GitHub repository.                      │ Github   │ 0.0.15  │
│ Github.CreateReplyForReviewComment     │ Create a reply to a review comment for a pull request.                    │ Github   │ 0.0.15  │
│ Github.CreateReviewComment             │ Create a review comment for a pull request in a GitHub repository.        │ Github   │ 0.0.15  │
│ Github.GetPullRequest                  │ Get details of a pull request in a GitHub repository.                     │ Github   │ 0.0.15  │
│ Github.GetRepository                   │ Get a repository.                                                         │ Github   │ 0.0.15  │
....

Authenticated Tools

Some MCP Servers may need authentication through an API key or OAuth, which can be configured in the Arcade Engine. To see the specific configuration requirements, you can checkout our Auth Providers.