How to vibe codeRobomotion

Prerequisites

• Node.jsfree

  Required for running the Next.js admin web console and the local TypeScript runner agent.

• GitHubfree

  Source control and CI/CD deployment pipeline.

AI coding tools

• Claude Code

  Best-in-class coding agent for multi-file scaffolding, connecting local runtimes, and building complex node canvas logic.

• Cursor

  Ideal for fine-tuning the Reactflow drag-and-drop canvas UI and debugging state synchronization.

Build guide

1. 01Scaffold Next.js Admin Console & Database

   Initialize the Next.js project with Tailwind CSS, better-auth, and Turso for managing flows, robot agents, and execution logs.

   Initialize a new Next.js project using Next.js 16, TypeScript, Tailwind CSS, and better-auth. Set up Turso (SQLite over HTTP) as the database with Drizzle ORM. Create database schemas for users, robots (development, production, application), flows (storing node graphs as JSON), and execution_logs (tracking run start, end, status, and step outputs). Implement a clean dark-mode dashboard layout with sidebar navigation for Flows, Robots, Credentials, and Settings. Write robust migration scripts and test database connectivity locally.

   ⧉copy promptopen in Claude CodeCopen in Codexopen in Cursor

2. 02Build the Visual Flow Designer Canvas

   Integrate React Flow to create a drag-and-drop automation canvas supporting custom nodes (RPA actions, HTTP requests, LLM agent loops).

   Install React Flow and build an interactive visual automation builder canvas inside the Next.js dashboard. Define custom node types: 'Action Node' (for HTTP/browser tasks), 'Logic Node' (if/else branching, loops), and 'Agent Node' (for LLM prompt execution). Implement drag-and-drop node creation from a sidebar palette, edge connection validation, and serialization of the canvas graph into a JSON schema that represents the automation flow sequence. Include a toolbar to save, test-run, and export flows.

   ⧉copy promptopen in Claude CodeCopen in Codexopen in Cursor

3. 03Develop the Local Desktop Runner Agent

   Build a standalone Node.js desktop runner service capable of polling the orchestration server and executing automation scripts locally.

   Create a standalone Node.js and TypeScript background worker service that acts as a local robot runner. The runner should authenticate with the central backend using a robot API key, poll a job queue for assigned flow executions, and download the flow JSON. Implement an execution engine that interprets the node graph step-by-step: executing HTTP requests, running JavaScript code snippets via isolated vm contexts, and logging step execution results back to the server in real-time.

   ⧉copy promptopen in Claude CodeCopen in Codexopen in Cursor

4. 04Implement the Secure Credential Proxy Vault

   Build an encrypted credential storage and proxy vault so automation nodes can authenticate against third-party APIs without exposing secrets.

   Implement a secure credential vault module using AES-256 encryption in Node.js. Create a Credential Proxy service that intercepts outgoing requests from executing robots, injects encrypted API keys or session cookies headers dynamically at runtime, and ensures that raw secrets are never passed into LLM prompt context windows or exposed in execution logs. Build UI screens in the Admin Console to manage and scope credentials to specific robots.

   ⧉copy promptopen in Claude CodeCopen in Codexopen in Cursor

5. 05Integrate Hybrid LLM Agent Loops

   Add AI assistant prompt generation and autonomous LLM agent execution nodes using the Vercel AI SDK and Anthropic API.

   Integrate the Vercel AI SDK and Anthropic API to support two AI capabilities: first, an 'AI Automation Builder' chat assistant in the Flow Designer that translates natural language prompts into valid React Flow JSON node graphs; second, an autonomous 'Agent Node' that uses Claude Sonnet with tool-calling capabilities to execute judgment-heavy steps within a workflow, making decisions and calling subflows as tools.

   ⧉copy promptopen in Claude CodeCopen in Codexopen in Cursor

6. 06Build Queue Orchestration & Concurrency Handling

   Implement task scheduling, parallel worker branching, and error retry logic across cloud and local runners.

   Build a robust concurrency and queue orchestration engine using Turso and background worker locks. Support multi-branching parallel execution where a workflow can split into multiple concurrent worker threads. Implement cron-based scheduling for automated job triggers, automatic error retries with exponential backoff, and dead-letter queues for failed runs. Add a live monitoring view in the Admin Console showing real-time job states and worker utilization.

   ⧉copy promptopen in Claude CodeCopen in Codexopen in Cursor

Cost vs paying for Robomotion