← Back to projects
05.2026
Rage AI: Multi-Model Agentic AI Platform
A full-stack agentic AI platform integrating OpenAI and Google Gemini with LangGraph state graphs, dynamic tool calling, and OAuth integrations.
Next.jsTypeScriptLangChainLangGraphPostgreSQLPrismaOpenAIGoogle Gemini
Introduction
Rage AI is an agentic AI workspace that bridges advanced Large Language Models with real-world tool execution. It allows users to orchestrate complex tasks across cloud services, run multi-turn agent workflows, and switch dynamically between frontier models.
Why I Built It
Simple chat interfaces fail when problems require external actions, long-running context, or multi-step reasoning. Rage AI was engineered to explore cyclic graph agent workflows (via LangGraph) that maintain explicit state, evaluate interim outputs, and call dynamically registered tools.
Key Capabilities
- Multi-Model Orchestration: Seamless switching between OpenAI GPT models and Google Gemini with streaming responses and persistent conversation threads.
- State Graph Agent Workflow: Built using LangGraph with conditional routing, self-correction, and structured tool invocation.
- Third-Party Integrations: OAuth 2.0 connectors for GitHub, Gmail, Google Drive, Google Calendar, and Notion with automatic token refresh handling.
- Granular Security: Organization and user-scoped data layers, encrypted OAuth credentials, soft deletion, and fine-grained permissions per tool.
- Modern User Experience: Built with Next.js 15, TypeScript, Tailwind CSS, and Markdown rendering with code highlighting and citation support.