Kokoron Meets Koclaw — Toward a More Complete AI Assistant
Kokoron Meets Koclaw — Toward a More Complete AI Assistant
AIKokoron started as a desktop companion — voice chat with a Live2D character. But I've been thinking bigger. What if the AI assistant could be more than a conversation partner? What if it could be a truly capable agent that works across platforms and handles real tasks autonomously?
Discovering OpenClaw and ZeroClaw
While researching AI assistant architectures, I came across two projects that changed my thinking:
- OpenClaw — a personal AI assistant that works across messaging platforms (WhatsApp, Telegram, Slack, Discord, etc.). It focuses on being available wherever you already communicate
- ZeroClaw — a Rust-native, ultra-lightweight AI agent runtime. It packs into a 3.4MB binary with under 5MB RAM usage, designed as infrastructure for agentic workflows where agents can be built once and run anywhere
These projects solve different problems than AIKokoron, but they share a common vision: AI that's truly integrated into your life, not just sitting in a browser tab.
The Koclaw Direction
Combining insights from these projects, I've started evolving AIKokoron into something I'm calling Koclaw — a more comprehensive AI assistant framework. The idea is to merge the best aspects of each approach:
- From AIKokoron — the interactive character experience, voice conversation, and MCP tool integration
- From OpenClaw — the multi-platform presence and messaging integration
- From ZeroClaw — the lightweight, deploy-anywhere agent runtime philosophy
The goal is an AI assistant that can:
- Live on your desktop as an interactive companion (existing AIKokoron functionality)
- Respond on messaging platforms — be available on your preferred chat apps
- Execute tasks autonomously — use tools, browse the web, manage files
- Run efficiently — lightweight enough to self-host without expensive infrastructure
Work in Progress
This is very much a work-in-progress. I'm currently:
- Refactoring the AIKokoron codebase to support a plugin architecture
- Designing the gateway layer that will handle multi-platform communication
- Experimenting with agent workflows for task automation
- Exploring how to integrate the lightweight runtime approach from ZeroClaw
The project hasn't been uploaded to GitHub yet — it's still in the experimental phase. But the architecture is taking shape, and I'm excited about where it's heading.
Why This Matters to Me
As someone studying software engineering, I believe the next frontier is AI that genuinely augments how we work. Not just answering questions, but actually doing things. The gap between "AI chatbot" and "AI agent" is where the interesting engineering challenges are.
I'll keep sharing progress here. If you're interested in AI assistant architectures, feel free to reach out via email or GitHub.