mcp-base 0.4.0

CLI tools for MCP (Model Context Protocol) server setup and management with OIDC support (Auth0, Dex, Keycloak)

MCP Base

CLI tools for MCP (Model Context Protocol) server setup and management. These tools help you configure OIDC authentication, create Kubernetes secrets, set up RBAC, and manage MCP server deployments.

Installation

From PyPI

# Install base package (OIDC tools only)
pip install mcp-base

# Install with Kubernetes support
pip install mcp-base[kubernetes]

# Install all optional dependencies
pip install mcp-base[all]

From Source

git clone https://github.com/your-org/mcp-base.git
cd mcp-base
pip install -e ".[all]"

Usage

After installation, the mcp-base command is available with the following subcommands:

mcp-base <command> [options]

Command	Description
setup-oidc	Set up OIDC provider (Auth0, Dex, Keycloak, etc.) for MCP authentication
create-secrets	Create Kubernetes secrets for MCP deployment
setup-rbac	Set up Kubernetes RBAC resources
add-user	Add users to allowed clients

Setting Up OIDC

Configure your OIDC provider for MCP authentication.

Supported providers: auth0, dex, keycloak, okta, generic

To see all provider options and examples:

mcp-base setup-oidc --help

To get help for a specific provider:

mcp-base setup-oidc --provider dex --help

1. Auth0 (Automated Setup)

The CLI automatically configures your Auth0 tenant:

# Set up Auth0 (first run)
mcp-base setup-oidc --provider auth0 \
    --domain your-tenant.auth0.com \
    --api-identifier https://mcp-server.example.com/mcp \
    --token YOUR_MGMT_TOKEN

# Subsequent runs (uses saved configuration)
mcp-base setup-oidc --provider auth0

# Force recreate clients (if secrets are lost)
mcp-base setup-oidc --provider auth0 --recreate-client

2. Dex, Keycloak, Okta, or Generic OIDC (Pre-configured)

For providers where you already have client credentials configured:

# Interactive mode (prompts for values)
mcp-base setup-oidc --provider dex

# Non-interactive mode
mcp-base setup-oidc --provider dex \
    --issuer https://dex.example.com \
    --audience https://mcp-server.example.com/mcp \
    --client-id YOUR_CLIENT_ID \
    --client-secret YOUR_CLIENT_SECRET

# Keycloak
mcp-base setup-oidc --provider keycloak \
    --issuer https://keycloak.example.com/realms/myrealm \
    --audience https://mcp-server.example.com/mcp \
    --client-id YOUR_CLIENT_ID \
    --client-secret YOUR_CLIENT_SECRET

# Okta
mcp-base setup-oidc --provider okta \
    --issuer https://your-org.okta.com \
    --audience https://mcp-server.example.com/mcp \
    --client-id YOUR_CLIENT_ID \
    --client-secret YOUR_CLIENT_SECRET

# Any generic OIDC provider
mcp-base setup-oidc --provider generic \
    --issuer https://your-idp.com \
    --audience https://mcp-server.example.com/mcp \
    --client-id YOUR_CLIENT_ID \
    --client-secret YOUR_CLIENT_SECRET

Required Redirect URLs

Configure these redirect URLs in your OIDC provider:

• MCP Server: https://mcp-server.example.com/auth/callback (replace with your actual server URL)
• Claude Desktop: https://claude.ai/api/mcp/auth_callback
• Local testing (optional): http://localhost:8888/callback, http://localhost:8889/callback

Endpoint resolution

When setup-oidc validates the issuer (the default), the authorization_endpoint, token_endpoint, and jwks_uri are read from the provider's .well-known/openid-configuration and saved verbatim to oidc-config.json. Pass --skip-validation to bypass the HTTP probe; in that case the endpoints are derived from provider-specific fallbacks:

Provider	Auth	Token	JWKS
keycloak	{issuer}/protocol/openid-connect/auth	{issuer}/protocol/openid-connect/token	{issuer}/protocol/openid-connect/certs
okta	{issuer}/v1/authorize	{issuer}/v1/token	{issuer}/v1/keys
dex, generic	{issuer}/auth	{issuer}/token	{issuer}/.well-known/jwks.json

Creating Kubernetes Secrets

Create secrets required for MCP server deployment:

# Create secrets in a namespace
mcp-base create-secrets --namespace default --release-name my-mcp-server

# Dry run to see what would be created
mcp-base create-secrets --namespace default --release-name my-mcp-server --dry-run

# Replace existing secrets
mcp-base create-secrets --namespace default --release-name my-mcp-server --force

Setting Up RBAC

Configure Kubernetes RBAC for MCP server:

# Cluster-wide permissions
mcp-base setup-rbac --namespace production --app-name my-mcp-server

# Namespace-scoped permissions
mcp-base setup-rbac --namespace production --app-name my-mcp-server --scope namespace

# Dry run
mcp-base setup-rbac --app-name my-mcp-server --dry-run

# Delete RBAC resources
mcp-base setup-rbac --app-name my-mcp-server --delete

Adding Users to Allowed Clients

Add users to the allowedClients array in your OIDC provider:

# Interactive mode
mcp-base add-user

# Non-interactive
mcp-base add-user --email [e-mail] --client-type both

Dependencies

Required

• requests>=2.28.0 - HTTP client for Auth0 API

Optional (Kubernetes support)

• kubernetes>=28.0.0 - Kubernetes Python client
• cryptography>=41.0.0 - For generating encryption keys

Install with: pip install mcp-base[kubernetes]

Development

Setup

# Clone the repository
git clone https://github.com/your-org/mcp-base.git
cd mcp-base

# Install in development mode with all dependencies
pip install -e ".[dev]"

Running Tests

pytest

Code Quality

# Format code
black src/

# Lint
ruff check src/

# Type checking
mypy src/

Publishing to PyPI

Prerequisites

1. Create an account on PyPI and Test PyPI
2. Install build tools:

   pip install build twine
   

3. Create API tokens and save them locally (git-ignored):

   echo "pypi-YOUR_TEST_TOKEN" > test.token && chmod 600 test.token
   echo "pypi-YOUR_PROD_TOKEN" > prod.token && chmod 600 prod.token
   

Makefile (preferred)

make build    # Clean + build only; artifacts in dist/
make test     # Run pytest
make dev      # Publish to Test PyPI using ./test.token
make prod     # Publish to production PyPI using ./prod.token (prompts "yes" to confirm)
make clean    # Remove dist/, build/, *.egg-info

make dev / make prod refuse to run if the expected token file is missing or empty.

Build the Package Manually

# Clean previous builds
rm -rf dist/ build/ *.egg-info src/*.egg-info

# Build source distribution and wheel
python -m build

This creates:

• dist/mcp_base-0.1.0.tar.gz (source distribution)
• dist/mcp_base-0.1.0-py3-none-any.whl (wheel)

Test on Test PyPI (Recommended)

# Upload to Test PyPI
python -m twine upload --repository testpypi dist/*

# Or with a token file
python -m twine upload --repository testpypi dist/* -u __token__ -p "$(cat ~/.testpypi-token)"

# Test installation from Test PyPI
pip install --index-url https://test.pypi.org/simple/ --extra-index-url https://pypi.org/simple/ mcp-base

Publish to PyPI

# Upload to PyPI
python -m twine upload dist/*

Using API Tokens (Recommended)

Instead of username/password, use API tokens:

1. Go to PyPI > Account Settings > API tokens
2. Create a token with scope "Entire account" or project-specific
3. Use the token directly:

   python -m twine upload dist/* -u __token__ -p pypi-YOUR_TOKEN_HERE
   

Using a Token File

For better security, store your token in a file and read it during upload:

# Store token in a file (do this once)
echo "pypi-YOUR_TOKEN_HERE" > ~/.pypi-token
chmod 600 ~/.pypi-token

# Upload using the token file
python -m twine upload dist/* -u __token__ -p "$(cat ~/.pypi-token)"

# Or using environment variables
TWINE_PASSWORD=$(cat ~/.pypi-token) python -m twine upload dist/*

Using .pypirc Configuration File

Or create ~/.pypirc:

[pypi]
username = __token__
password = pypi-YOUR_TOKEN_HERE

[testpypi]
username = __token__
password = pypi-YOUR_TEST_TOKEN_HERE

Automated Publishing with GitHub Actions

Create .github/workflows/publish.yml:

name: Publish to PyPI

on:
  release:
    types: [published]

jobs:
  publish:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

      - name: Set up Python
        uses: actions/setup-python@v5
        with:
          python-version: '3.11'

      - name: Install build dependencies
        run: pip install build twine

      - name: Build package
        run: python -m build

      - name: Publish to PyPI
        env:
          TWINE_USERNAME: __token__
          TWINE_PASSWORD: ${{ secrets.PYPI_API_TOKEN }}
        run: python -m twine upload dist/*

Add your PyPI API token as a repository secret named PYPI_API_TOKEN.

License

MIT License - see LICENSE file for details.