MCP Server
Connect AI assistants to Portabase via the Model Context Protocol (MCP) server.
The Portabase MCP server exposes your dashboard over the Model Context Protocol, allowing AI assistants (Claude, Cursor, Windsurf, etc.) to manage databases, agents, and backups through natural language.
Prerequisites
- Portabase dashboard running with both
API_ENABLED=trueandMCP_ENABLED=true - An API token (see API Introduction)
- Node.js 18+ on the machine running your AI assistant
Enable MCP
Set both environment variables before starting your dashboard:
API_ENABLED=true
MCP_ENABLED=trueAPI_ENABLED=true: enables all API routes under/api/v1MCP_ENABLED=true: enables the MCP server at/api/v1/mcp
Connection
Add the following to your AI assistant's MCP configuration, replacing the URL and API key with your own:
Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):
{
"mcpServers": {
"portabase": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://your-dashboard.example.com/api/v1/mcp",
"--header",
"x-api-key: YOUR_API_TOKEN"
]
}
}
}Open Settings → MCP Servers and add:
{
"portabase": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://your-dashboard.example.com/api/v1/mcp",
"--header",
"x-api-key: YOUR_API_TOKEN"
]
}
}Any MCP-compatible client that accepts a JSON config:
{
"mcpServers": {
"portabase": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://your-dashboard.example.com/api/v1/mcp",
"--header",
"x-api-key: YOUR_API_TOKEN"
]
}
}
}Never commit your API token to version control. Use your environment's secret management to inject it where possible.
Verify the connection
Restart your AI assistant. Ask it:
"List my Portabase databases"
A successful response confirms the MCP server is connected.
Available Tools
See the Tools reference for the full list of operations.
Last updated on