BanterAI | Devpost

BanertAI RelationShip
BanertAI Chat
BanertAI Settings

Inspiration

Modern software development is no longer limited by coding — it is limited by communication.

Engineers spend more time discussing solutions, clarifying requirements, aligning decisions, and coordinating across teammates than actually writing code. Yet collaboration tools remain fragmented: chat apps, documents, repos, and AI assistants all live in separate spaces.

We asked:

What if collaboration, communication, and AI assistance could exist directly inside the same repo context?

BanterAI was built to unify team collaboration and AI into a single workflow layer.

What it does

BanterAI is an AI-powered collaboration system designed for engineering teams working under the same repository.

Its core features include:

Real-time Repo Chat (“Trash Talk”) Teammates can communicate instantly inside the repo context — from friendly banter to intense technical debates.

AI + MCP Integration Through Model Context Protocol (MCP), users can instruct AI to act operationally. For example: “Send the solution we discussed to BoBo.” The AI synthesizes a structured, human-readable summary and delivers it automatically.

Persona-Based Messaging Modes AI tone can be customized per teammate — toxic, neutral, sweet, or playful — enabling more natural social dynamics in collaboration.

Built-in Translation Teams can collaborate across languages with one-click translation or AI-driven multilingual messaging.

Additional features include private messaging, image sharing, and AI participation in discussions.

How we built it

BanterAI combines multiple collaboration and AI technologies:

Real-time messaging powered by Ably infrastructure
Repo-bound identity mapping via Git username and repo ID
MCP (Model Context Protocol) for AI action execution
LLM synthesis for summarization and message generation
Tone modulation layers for persona customization
Translation integrations for cross-language collaboration

The system is designed to embed AI directly into team communication flows rather than operate as a separate assistant.

Challenges we ran into

One major challenge was context binding.

Ensuring that AI understands discussion context, solution decisions, and recipient targets inside a fast-moving chat environment required careful orchestration between MCP triggers and LLM summarization.

Another challenge was balancing tone customization with clarity — allowing expressive team culture (including humor or trash talk) without degrading message usefulness.

Accomplishments that we're proud of

Built a real-time repo collaboration chat system
Enabled AI to execute communication tasks via MCP
Implemented persona-based tone modulation
Integrated multilingual collaboration features
Created a workflow where AI participates as a team member — not just a tool

What we learned

We learned that collaboration friction often outweighs technical difficulty.

When AI is embedded directly into team discussion loops — summarizing, relaying, translating, and contextualizing — productivity increases without forcing teams to change platforms.

Human dynamics, humor, and tone also play a significant role in engineering collaboration, and AI must adapt to them rather than suppress them.