ChatterBoxx

A software that allows you to converse in real time with multiple versions of yourself, ranging from Shakespearean poet to cyberpunk hacker. AI-powered, MCP-synchronized, and fun-driven.

Comment

Inspiration

We were inspired by the concept of the multiverse and the question: "What if you could talk to different versions of yourself from alternate realities?" This idea, combined with the power of large language models, led us to create ChatterBoxx — a multiverse chat simulator where your alternate selves come alive with unique personas, behaviors, and storylines. We wanted to blend narrative creativity, AI, and real-time systems into one chaotic, fun, and meaningful experience.

What it does

ChatterBoxx lets users chat with multiple AI-generated versions of themselves, each from a different timeline or reality. These alternate selves have distinct personas (e.g., pirate version, robotic version, poetic version), generated using prompt-engineered LLMs. The MCP server coordinates and synchronizes all messages, persona roles, and “glitch events” in real time. Users can view, interact with, or spectate timeline collisions — including unexpected persona swaps and merged conversations.

How we built it

  • Frontend: React with Next.js and TailwindCSS for the chat interface. Framer Motion is used to add glitch/timeline effects and character transitions.
  • Backend: Node.js server integrated with Python-based prompt pipeline to coordinate with LLM APIs (OpenAI and Perplexity).
  • MCP Server: Orchestrates real-time user sessions, synchronizes multiverse chat states, tracks active personas, and manages glitch events.
  • Persona Engine: Extracts a base persona profile by analyzing chat history, then applies timeline transformations to generate alternate selves.
  • Memory Store: Stores historical context per persona to ensure consistency across sessions.

Challenges we ran into

  • Creating believable and consistent personas using LLMs across varied timelines
  • Managing multiple asynchronous LLM responses in a real-time chat environment
  • Designing glitch mechanics that were chaotic but controlled and recoverable
  • Keeping latency low while processing complex prompt injections
  • Building a UX that made sense while being visually representative of the multiverse concept

Accomplishments that we're proud of

  • Built a fully working real-time multiverse chat experience within the hackathon timeframe
  • Developed a persona transformation pipeline that consistently produces entertaining and coherent AI personalities
  • Used the MCP server effectively to orchestrate timeline events and maintain real-time communication among personas
  • Designed an interface that captures the imagination and generates social-worthy, shareable content

What we learned

  • How to design and prompt-engineer role-specific LLM agents at scale
  • Using the MCP server for complex, dynamic state synchronization across clients
  • Real-time coordination of multi-agent conversational systems
  • Balancing creativity and control in generative AI experiences
  • Techniques for embedding, extracting, and using chat memory to build base personas

What's next for ChatterBoxx

  • Persona Marketplace: Let users design and share custom alternate selves
  • Social Integration: Connect ChatterBoxx personas with real apps like WhatsApp, Instagram DMs, etc.
  • Timeline Summarizer: Generate summary comics or stories from chaotic chat sessions using LLMs
  • Streaming Mode: Let creators live-stream multiverse interactions
  • Voice Integration: Enable spoken interactions and dynamic voice synthesis per persona
  • Multiverse Leaderboard: Track the most chaotic, funniest, or profound alternate conversations

ChatterBoxx began as an experiment in creative chaos, but we believe it has real potential to become a new genre of AI-powered social entertainment.

Built With

Share this project:

Updates