Merc-bot

To turn Merc-bot into a truly extensible platform, we exposed the full cross-platform "event bus" as a structured API and SDKs.

This now lets other developers hook into the signals (alerts, commands, subscriptions) sent and received by our architecture. New projects, like dynamic 3D environments or custom alerts, can come to life very easily.

Merc-bot evolves to have its own on-screen personality. By feeding Twitch and Discord chat streams into an agentic intelligence architecture, the bot can infer context and respond to the community in a human way.

Using OpenAI models for reasoning and phrasing, combined with the powerful ElevenLabs voice synthesis engine, lets each bot speak with a voice unique to the streamer.

The success of the initial prototype encouraged us to grow the solution into a real service. The major 2025 overhaul was the move to a Multi-Tenant architecture.

Now Merc-bot can host many streamers on the same application infrastructure. Each one manages their own bot, with full, secure isolation of data, voice tokens, and respective API keys.

Engagement automation is one of a streamer's biggest challenges for keeping a community loyal. That is the heart of the Discord integration: unifying the live "stream" space and the asynchronous "server" space.

When a live starts, Merc-bot synchronizes the event without friction and updates users through Discord, with automatic real-time post editing of thumbnails and current games.

The genesis of the project. Connecting the bot directly to the streamer's lifeblood: their chat and Twitch event streams via EventSub and WebSockets.

This phase laid the architectural foundations for reading. We built a robust event bus that connects our own 3D application (3D overlay) directly to Twitch actions, to gamify viewer interactions.

In this dynamic environment, each viewer has their own avatar ('Fromage') that floats interactively in the stream space. Their messages are displayed physically there, letting them literally become part of the broadcast.