Welcome to My Digital Lab
Built on the ZeroClaw framework.
Why I built this
I wanted to build something I actually own — not just use someone else's AI tool. This is what I built and what went wrong along the way.
Current lead agent: Nova
Nova is my Content Intelligence Officer — a YouTube specialist that finds the most relevant videos on my chosen topics and delivers tight summaries.
- Searches YouTube via the Data API v3 through a custom MCP server
- Delivers top 3 videos with 3-bullet summaries to Telegram
- Deep-dive analysis on request (reply with a number)
- Topics managed via
/topicscommands
How it works
ZeroClaw acts as the manager — it receives commands via Telegram, delegates YouTube work to Nova, then renders thumbnails and presents results. Nova focuses purely on content analysis.
The YouTube pipeline: MCP server → Python fetcher → YouTube Data API v3.
Navigation
- Nova Identity — who Nova is
- Nova Soul — core values and rules
- Nova Skill — the briefing workflow
- Nova Style — output format
- Manager Soul — how ZeroClaw delegates
- YouTube Fetcher — the Python pipeline
- MCP Server — the tool server
- Configuration — sanitised config
- Troubleshooting — what broke and why
- Showcase — architecture and results
Key numbers
| Metric | Value |
|---|---|
| Framework | ZeroClaw v0.3.4 |
| VPS | Hetzner, 2 vCPU, 3.7GB RAM |
| Model | Claude Haiku 4.5 (via OpenRouter) |
| Lead agent | Nova — YouTube Content Intelligence |
| Topics | AI agents, Super Individual business, Claude AI |
| Cost limits | $5/day, $20/month |
| Security | Workspace sandbox, Telegram allowlist, audit logging |
| Monthly cost | ~$6 (VPS) + ~$5-15 (API) |