Skip to content

MCP Integration

DevAll includes a built-in MCP (Model Context Protocol) server that allows AI assistants like Claude to control your entire development environment through natural language.

The Model Context Protocol is a standard for connecting AI assistants to external tools and data sources. DevAll’s MCP server exposes tools that let AI assistants:

  • Start and stop services
  • View service status and logs
  • Restart processes
  • Modify configuration
  • Monitor system resources
  1. Set up the MCP server:
Terminal window
npx devall mcp setup

This will:

  • Build the MCP server
  • Configure Claude Desktop automatically
  • Add DevAll to your Claude Desktop settings
  1. Start the DevAll dashboard:
Terminal window
devall
  1. Open Claude Desktop and you’re ready! Claude can now control your dev environment.

If automatic setup doesn’t work, configure Claude Desktop manually:

Terminal window
cd mcp
npm install
npm run build

Edit your Claude Desktop config file:

macOS/Linux:

~/Library/Application Support/Claude/claude_desktop_config.json

Windows:

%APPDATA%\Claude\claude_desktop_config.json

Add the DevAll MCP server:

{
"mcpServers": {
"devall": {
"command": "node",
"args": ["/path/to/devall-core/mcp/bin/index.js"]
}
}
}

Close and reopen Claude Desktop to load the MCP server.

Once configured, you can control DevAll naturally:

“Start the frontend service”

“Launch all autostart services”

“Start the API and database”

“Stop the backend server”

“Shut down all services”

“What services are running?”

“Show me the status of all services”

“Is the frontend service up?”

“Show me the last 50 lines of the API logs”

“What errors are in the backend logs?”

“Restart the frontend”

“Restart all services”

“What services are configured?”

“Add a new service for the authentication server”

DevAll exposes these tools to AI assistants:

Lists all configured services with their status.

Output:

{
"services": [
{
"name": "Frontend",
"status": "running",
"port": 3000,
"pid": 12345
}
]
}

Gets detailed status for a specific service.

Parameters:

  • name (string): Service name

Output:

{
"name": "Frontend",
"status": "running",
"port": 3000,
"pid": 12345,
"uptime": "5m 23s"
}

Starts a service.

Parameters:

  • name (string): Service name

Stops a running service.

Parameters:

  • name (string): Service name

Restarts a service (stops then starts).

Parameters:

  • name (string): Service name

Gets recent logs for a service.

Parameters:

  • name (string): Service name
  • lines (number, optional): Number of lines (default: 100)

Output:

{
"logs": [
"Server listening on port 3000",
"GET / 200 15ms",
"GET /api/users 200 45ms"
]
}

Clears logs for a service.

Parameters:

  • name (string): Service name

Gets the current DevAll configuration.

Output:

{
"dashboard": {
"port": 7777
},
"services": [...]
}

You can run the MCP server independently of Claude Desktop:

Terminal window
npx devall mcp start
Terminal window
npx devall mcp start --sse --port 3001

Access at: http://localhost:3001/sse

The MCP Inspector is a tool for testing MCP servers:

Terminal window
npm run mcp:inspector

This opens a web interface where you can:

  • Test all available tools
  • View request/response payloads
  • Debug MCP server behavior

Watch for changes and auto-reload:

Terminal window
cd mcp
npm run dev

The MCP server currently has no authentication. Ensure:

  1. DevAll dashboard is only accessible on localhost
  2. MCP server only accepts local connections
  3. Don’t expose ports externally

The MCP server has full control over configured services. It can:

  • Start/stop any process
  • View all logs
  • Modify configuration

Only use with trusted AI assistants in secure environments.

  1. Check Claude Desktop config is correct:

    Terminal window
    cat ~/Library/Application\ Support/Claude/claude_desktop_config.json
  2. Verify the path to index.js is absolute

  3. Check MCP server builds successfully:

    Terminal window
    cd mcp && npm run build
  4. Restart Claude Desktop completely

The MCP server requires the DevAll dashboard to be running:

Terminal window
# Terminal 1: Start dashboard
devall
# Terminal 2: Use Claude Desktop
# (MCP tools now work)
  1. Check DevAll dashboard is accessible:

    Terminal window
    curl http://localhost:7777/api/health
  2. View MCP server logs (if running standalone):

    Terminal window
    npx devall mcp start
    # Check stderr output
  3. Test with MCP Inspector:

    Terminal window
    npm run mcp:inspector

“I need to work on the frontend. Start the frontend service and the API gateway.”

AI will:

  1. Call start_service for “Frontend”
  2. Call start_service for “API Gateway”
  3. Confirm both are running

“The backend is behaving strangely. Show me the last 100 log lines.”

AI will:

  1. Call get_service_logs with lines: 100
  2. Analyze the logs
  3. Suggest fixes based on errors

“I just updated the environment variables. Restart all services.”

AI will:

  1. Call list_services to get all services
  2. Call restart_service for each one
  3. Verify they’re all running

“Are all my services running? What’s their status?”

AI will:

  1. Call list_services
  2. Present a summary of service states
  3. Alert you to any stopped/crashed services

You can extend DevAll’s MCP server with custom tools:

  1. Edit mcp/src/index.js
  2. Add your tool to the tools list
  3. Implement the tool handler
  4. Rebuild: npm run build

Claude Desktop can use multiple MCP servers simultaneously. Combine DevAll with:

  • File system tools: Read/write code
  • Git tools: Commit changes
  • Database tools: Query data
  • Slack tools: Send notifications

Example config:

{
"mcpServers": {
"devall": {
"command": "node",
"args": ["/path/to/devall/mcp/bin/index.js"]
},
"filesystem": {
"command": "npx",
"args": ["@modelcontextprotocol/server-filesystem"]
}
}
}