A community-driven registry for Claude, Cursor, Windsurf, Cline & more. Not affiliated with Anthropic.
Are you the author? Sign in to claim
Compose, manage, and run MCP servers as Docker containers. With a Unified API gateway built in.
Docker Compose for Model Context Protocol servers. Run multiple MCP servers with unified configuration, HTTP proxy, and container orchestration.

git clone https://github.com/phildougherty/mcp-compose.git
cd mcp-compose
make build
# Add to PATH
sudo cp build/mcp-compose /usr/local/bin/
# OR
export PATH="$PWD/build:$PATH"
mcp-compose init
Follow the prompts to create your configuration.
Create mcp-compose.yaml:
version: '1'
servers:
filesystem:
protocol: stdio
command: npx
args:
- "-y"
- "@modelcontextprotocol/server-filesystem"
- "/tmp"
capabilities: [resources, tools]
Start servers:
mcp-compose up
Start the HTTP proxy:
mcp-compose proxy --port 9876
Test it:
curl http://localhost:9876/api/servers
version: '1'
servers:
filesystem:
protocol: stdio
command: npx
args:
- "-y"
- "@modelcontextprotocol/server-filesystem"
- "${HOME}"
capabilities: [resources, tools]
volumes:
- "${HOME}:${HOME}:ro"
memory:
protocol: stdio
command: npx
args:
- "-y"
- "@modelcontextprotocol/server-memory"
capabilities: [resources, tools]
Add Brave Search (requires API key from https://brave.com/search/api/):
brave-search:
protocol: stdio
command: npx
args:
- "-y"
- "@modelcontextprotocol/server-brave-search"
capabilities: [tools]
env:
BRAVE_API_KEY: "${BRAVE_API_KEY}"
export BRAVE_API_KEY="your-api-key"
mcp-compose up
See examples/ for more configurations.
mcp-compose up # Start all services
mcp-compose up filesystem # Start specific service
mcp-compose down # Stop all services
mcp-compose down filesystem # Stop specific service
mcp-compose restart # Restart all services
mcp-compose ps # List service status
mcp-compose logs filesystem # View logs
System services are infrastructure components (proxy, dashboard, task-scheduler, memory).
mcp-compose system up # Start system services
mcp-compose system down # Stop system services
mcp-compose system ps # List system services
mcp-compose system status # Health overview
mcp-compose system logs proxy # View system logs
mcp-compose proxy --port 9876 # Start proxy
mcp-compose proxy --api-key $(openssl rand -hex 32) # With authentication
mcp-compose dashboard # Start web dashboard
# Access at http://localhost:3111
mcp-compose init # Interactive setup wizard
mcp-compose validate # Validate config file
mcp-compose create-config --type claude # Generate client config
version: '1'
# Optional: proxy authentication
proxy_auth:
enabled: true
api_key: "${MCP_API_KEY}"
servers:
my-server:
protocol: string # stdio, http, sse
command: string # Executable (e.g., npx, python, node)
args: [string] # Command arguments
capabilities: [string] # tools, resources, prompts, sampling
env: # Environment variables
KEY: value
volumes: # Volume mounts
- "host:container:mode"
STDIO (most common for NPM packages):
servers:
my-server:
protocol: stdio
command: npx
args: ["-y", "@modelcontextprotocol/server-filesystem", "/tmp"]
HTTP:
servers:
my-server:
protocol: http
http_port: 8080
SSE (Server-Sent Events):
servers:
my-server:
protocol: sse
http_port: 8080
sse_path: /events
volumes:
- "/host/path:/container/path:ro" # Read-only
- "/host/path:/container/path:rw" # Read-write
- "${HOME}/code:/workspace:ro" # Environment variables
In your config:
env:
API_KEY: "${MY_API_KEY}"
In your shell:
export MY_API_KEY="secret"
mcp-compose up
Never commit secrets to your config file. Use environment variables:
# Generate secure keys
export MCP_API_KEY=$(openssl rand -hex 32)
export OAUTH_CLIENT_SECRET=$(openssl rand -hex 32)
Reference in config:
proxy_auth:
api_key: "${MCP_API_KEY}"
Generate config:
mcp-compose create-config --type claude --output ./claude-config
Copy contents to Claude Desktop settings.
# List tools
curl -X POST http://localhost:9876/filesystem \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'
# Call tool
curl -X POST http://localhost:9876/filesystem \
-H "Content-Type: application/json" \
-d '{
"jsonrpc":"2.0",
"id":1,
"method":"tools/call",
"params":{
"name":"read_file",
"arguments":{"path":"/tmp/test.txt"}
}
}'
mcp-compose ps
mcp-compose system status
mcp-compose logs filesystem
mcp-compose system logs proxy
mcp-compose validate
Port already in use:
lsof -i :9876
mcp-compose proxy --port 9877 # Use different port
Container not starting:
mcp-compose logs my-server # Check logs
docker ps -a # Check container status
Authentication errors:
echo $MCP_API_KEY # Verify key is set
┌──────────────────────────────┐
│ MCP-Compose Proxy │
│ (HTTP API + Authentication) │
└──────────────┬───────────────┘
│
┌──────┼──────┐
│ │ │
┌───▼──┐ ┌─▼──┐ ┌▼───┐
│ FS │ │ Mem│ │Search│
│STDIO │ │HTTP│ │ SSE │
└──────┘ └────┘ └──────┘
GNU Affero General Public License v3.0 - see LICENSE.
MCP server integration for DaVinci Resolve Studio
Run Claude Code as an MCP server so any agent can delegate coding tasks to it
Browser automation using accessibility snapshots instead of screenshots
A Jetbrains IDE IntelliJ plugin aimed to provide coding agents the ability to leverage intelliJ's indexing of the codeba