Four AI agents working together โ€” coding, monitoring, research, and knowledge retrieval. Each has a distinct role and personality.

๐Ÿ’ป
Claude Code
Primary coding agent. Architecture, implementation, debugging, code review, git workflow. Runs on Anthropic's Claude Opus. Has 25 language rules, 18 slash commands, 16 specialist sub-agents.
JS/TS Rust C Python Go C++
๐Ÿ‘๏ธ
Argus
Infrastructure watchdog and research processor. Runs on Hermes Agent with local Ollama inference (cogito:14b, 48% GPU / 52% CPU). Monitors Fiber nodes, processes research tasks, alerts on failures.
Hermes Ollama Cron Autonomous
๐Ÿ“ก
Shannon
RAG agent for exploratory research. Named after the smallest unit of CKB (and Claude Shannon). Temperature 0.7 for creative/broad queries. Queries the shared knowledge base of 219+ research findings.
RAG Exploratory 0.7T
๐Ÿ”ง
Kernel
RAG agent for code-focused queries. Temperature 0.2 for precise, deterministic answers. Same knowledge base as Shannon but optimized for technical accuracy over breadth.
RAG Code-focused 0.2T

How They Work Together

Claude Code submits research tasks to Argus's inbox via SSH. Argus processes them with Gemini Flash (Ollama fallback), writes findings, and ingests them into the shared RAG. Shannon and Kernel query those findings. All agents share the same knowledge base โ€” what one learns, all can access.

โš™๏ธ
OpenClaw Setup
Installing and configuring OpenClaw agents
๐Ÿง 
Local LLM Inference
Running Ollama models locally, GPU offloading, model selection for your hardware
๐Ÿ“–
RAG Systems
Building retrieval-augmented generation with Obsidian vault sync and agent-shared knowledge bases
๐Ÿค–
Agent Development
Creating and deploying agentic systems โ€” from single Claude Code agent to multi-agent orchestration
๐Ÿ”’
Safety Hooks
Protecting your system from AI mistakes. Pre-tool-use hooks that block destructive commands, .env writes, force pushes. Coming soon.
๐Ÿ“
CLAUDE.md Guide
How to write effective CLAUDE.md files that give your agent project context, build commands, and conventions. Coming soon.
๐Ÿงฉ
Multi-Agent Setup
Running Claude Code + Hermes + RAG agents together. Research inboxes, shared knowledge, autonomous monitoring. Coming soon.
๐Ÿ“Š
Benchmarking with AI
Using AI agents to design, run, and analyse performance benchmarks. Real example: CKB light client RocksDB vs SQLite. Coming soon.

Wyltek Studio โ€” AI-powered creative tools for content creation and promotion.

๐Ÿ–ผ๏ธ
Image Generation
Local Stable Diffusion via ComfyUI. Custom workflows for project artwork, social media, documentation visuals. Models stored on dedicated NVMe.
ComfyUI SDXL LoRA
๐ŸŽค
Text-to-Speech
Voice synthesis for video narration, tutorials, and announcements. Local inference via Ollama TTS models.
TTS Local
๐ŸŽต
Music Generation
AI-generated background music and sound effects for videos and game demos. AudioCraft integration.
AudioCraft
๐ŸŽฌ
Video Timeline
Compose image, audio, TTS, and music into video timelines for demos, tutorials, and social content. FFmpeg-based pipeline.
FFmpeg Pipeline

Automated research pipeline โ€” agents submit topics, Argus processes them, findings feed into the shared knowledge base.

๐Ÿ“Š
Browse Findings
219+ structured research findings. CKB, embedded systems, AI, quantization, payment channels, wallet architecture. Filterable and searchable.
๐Ÿ”„
Research Pipeline
Any agent submits a task (markdown file) to Argus's inbox. Argus fetches seed URLs, analyses with Gemini Flash (Ollama fallback), writes findings, ingests to RAG. Every 10 minutes.
๐Ÿ“ก
Auto-Publish
Findings auto-sync to this website every 6 hours. Also pushed to Obsidian vault for offline access. Git-backed, versioned.
๐Ÿงช
Recent: CKBull Analysis
Performance analysis of CKBull wallet โ€” found 5.6x speedup opportunity by switching from cell enumeration to get_cells_capacity. Benchmarked against their live mainnet endpoint.