Skip to main content
The GoRules CLI (@gorules/cli) is an open-source command-line tool that bridges GoRules BRMS projects to your local development environment. It exposes an MCP (Model Context Protocol) server so AI-powered editors like Claude Code, Cursor, and Windsurf can read, evaluate, and interact with your decision logic directly. GitHub: github.com/gorules/cli

​
Installation

# npm
npm install -g @gorules/cli

# Or run directly with npx
npx @gorules/cli mcp start

​
Quick start

Start the MCP bridge: 
gorules mcp start
This launches a local server on port 41919 that:
  1. Exposes an MCP endpoint for AI tools to discover and call GoRules tools
  2. Exposes REST endpoints for evaluating decisions and fetching decision files
  3. Opens a WebSocket connection that the GoRules browser editor connects to
Once running, open your GoRules project in the browser and click Connect MCP to link the editor to your local CLI.

​
Commands

​
gorules mcp start

Start the MCP bridge server.
FlagAliasDefaultDescription
--port-p41919Server port
--host-hlocalhostServer host
--url-u-GoRules server URL
--open-falseOpen browser on start

​
REST API

When the bridge is running, it exposes HTTP endpoints for local development.

​
Evaluate a decision

POST http://localhost:41919/evaluate/{filePath}

{
  "context": {
    "customer": { "tier": "premium" },
    "orderTotal": 150
  },
  "trace": false
}
Response: 
{
  "result": { "discount": 0.15, "freeShipping": true }
}
Optional body fields: trace (boolean), maxDepth (number).

​
Retrieve a decision file

GET http://localhost:41919/file/{filePath}
Returns the raw decision graph JSON. You can use this as a loader for the ZEN Engine: 
import { ZenEngine } from "@gorules/zen-engine";

const engine = new ZenEngine({
  loader: async (key) => {
    const res = await fetch(`http://localhost:41919/file/${key}`);
    return res.json();
  },
});

const result = await engine.evaluate("my-decision", {
  customer: { tier: "premium" },
  orderTotal: 150,
});

​
How it works

The CLI acts as a bridge between AI tools and the GoRules browser editor:
  1. The CLI starts and generates a connection token
  2. The GoRules browser editor connects via WebSocket using the token
  3. The editor sends a tool manifest - the list of available tools for that project
  4. AI tools discover these tools via MCP and invoke them
  5. The CLI forwards tool calls to the browser, which executes them and returns results
The CLI is stateless - it doesn’t store credentials or project data. All tool execution happens in the browser editor, which has an authenticated session with the GoRules platform.

​
License

MIT