@restforgejs/mcp-server

MCP (Model Context Protocol) server for the RESTForge framework. Exposes RESTForge capabilities to AI Agents (Claude Desktop, Cursor, Claude CLI, and other MCP clients) so agents can operate RESTForge through natural language without manually invoking CLI commands.

Requirements

• Node.js >= 18
• npm >= 9
• For full setup workflow: PostgreSQL / MySQL / Oracle / SQLite, RESTForge license key

Installation

npm install -g @restforgejs/mcp-server

After installation, the restforge-mcp command is available in PATH.

Quick Start

1. Verify Install

echo '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' | restforge-mcp

Output should list 29 tools across the health_*, setup_*, codegen_*, and runtime_* domains.

2. Register with MCP Client

Claude CLI (user scope, applies to all projects):

claude mcp add --transport stdio --scope user restforge -- restforge-mcp

Cursor (.cursor/mcp.json in project root):

{
  "mcpServers": {
    "restforge": {
      "command": "restforge-mcp"
    }
  }
}

Claude Desktop (claude_desktop_config.json):

{
  "mcpServers": {
    "restforge": {
      "command": "restforge-mcp"
    }
  }
}

3. Use via Natural Language

In your AI client chat, type prompts like:

Setup a new RESTForge project at d:/projects/api-test with PostgreSQL on localhost:5432, license XXXX-XXXX-XXXX-XXXX

Generate a CRUD endpoint for the customer table

Run my RESTForge server (the agent generates a launcher script for the user to execute)

The agent orchestrates the appropriate tools to fulfill the request end-to-end.

Available Tools

29 tools organized by domain. AI agents call these via the MCP protocol; end users do not invoke them directly.

Health Domain (1 tool)

| Tool | Description | |------|-------------| | health_ping | Smoke test MCP transport. Returns pong + ISO timestamp + server version |

Setup Domain (9 tools)

| Tool | Description | |------|-------------| | setup_create_folder | Create a new project folder for RESTForge | | setup_install_package | Install @restforgejs/platform into the project's node_modules via npm | | setup_init_config | Generate skeleton config and sample payloads via restforge init | | setup_write_env | Write config/db-connection.env with license, server, and database settings | | setup_read_env | Read current values from config/db-connection.env | | setup_update_env | Update individual fields in config/db-connection.env | | setup_validate_config | Validate license and connections to database, redis, and kafka | | setup_get_config_schema | Get JSON schema of all 63 parameters available in db-connection.env | | setup_get_init_template | Get raw db-connection.env template content |

Codegen Domain (13 tools)

| Tool | Description | |------|-------------| | codegen_list_tables | List all tables in the project's database (live introspection) | | codegen_describe_table | Describe columns, primary key, and foreign keys of a specific table | | codegen_generate_payload | Generate payload JSON from a database table | | codegen_validate_payload | Validate payload JSON structure and constraints | | codegen_validate_dashboard_payload | Validate dashboard payload structure | | codegen_diff_payload | Diff payload JSON against the database schema | | codegen_sync_payload | Sync payload JSON with the database schema | | codegen_create_endpoint | Scaffold an endpoint module from a payload spec | | codegen_create_dashboard | Scaffold a dashboard module from a payload spec | | codegen_validate_sql | Validate a SELECT or WITH (CTE) SQL statement via EXPLAIN against the live database | | codegen_get_field_validation_catalog | Get the field validation catalog (for grounding payload constraints) | | codegen_get_query_declarative_catalog | Get the query declarative catalog (for grounding query JSON) | | codegen_get_dashboard_catalog | Get the dashboard widget catalog (for grounding dashboard config) |

Runtime Domain (6 tools)

| Tool | Description | |------|-------------| | runtime_detect_project | Scan src/modules/*.js to list project names | | runtime_detect_config | Scan config/*.env to list available config files | | runtime_validate_preflight | Validate config + check PID file + check port availability before launch | | runtime_check_launcher_exists | Check if launcher files (server-start.bat/.sh, ecosystem.config.js) exist in the project root | | runtime_generate_launcher | Generate server-start.bat/.sh + server-stop.bat/.sh (and ecosystem.config.js for PM2 mode) | | runtime_check_status | Detect if the server is running (host or PM2 mode) with optional HTTP health probe |

Runtime principle: AI agents never start, stop, or restart the server directly. The runtime tools only generate launcher scripts that the user executes themselves, so the running server lives independently of the AI session.

Compatibility

This MCP server works with any MCP client that supports the stdio transport, including but not limited to:

• Claude Desktop
• Claude CLI (Claude Code)
• Cursor
• Windsurf
• Cline (VS Code extension)
• Continue (VS Code/JetBrains extension)
• Zed

The model used (Claude, GPT, Gemini, etc.) depends on the client configuration. Tool selection accuracy is best with frontier models that have mature tool-calling support.

Repository

• Source: https://github.com/restforge-dev/mcp-server
• Issues: https://github.com/restforge-dev/mcp-server/issues

License

MIT — see LICENSE.md.