nvk
nvk is an individual open-source developer who created LLM Wiki, a local-first system for compiling source-grounded knowledge bases for AI agents. It combines parallel research, source ingestion, wiki compilation, querying, session memory, audits, and artifact generation into portable agent plugins and Markdown workflows.
AI Tools by nvk
(1)
LLM Wiki
Open Source LLM Research Wiki
Discussions
No discussions yet
Be the first to start a discussion about nvk
Latest News
Products & Services
Open-source local-first knowledge-base and agent workflow system. It is distributed as a Claude Code plugin, an OpenAI Codex plugin, an OpenCode instruction package, and portable AGENTS.md instructions; it is also Obsidian-compatible.
A companion public-source browser extension workflow that gives a local agent bounded, visible control of only an HTTPS tab through fixed MCP tools; grants end on stop, close, or cross-origin navigation.
A declared-route integration layer for external tools, with scoped resources, plan-bound approval, read-back verification, and adapter-owned workflow guides.
Market Position
LLM Wiki is positioned as a local-first, open-source alternative to chat-only research and hosted RAG/knowledge-base products: it keeps raw sources and Markdown artifacts under the user's control, emphasizes provenance, deterministic maintenance, privacy boundaries, and multi-runtime agent integration. Comparable community approaches include Karpathy's original LLM-wiki pattern and other Claude Code knowledge-base plugins such as rvk7895/llm-knowledge-bases; the repository also recommends tobi/qmd as a local search engine when a wiki grows beyond roughly 100 articles.
Leadership
Founders
nvk (Rodolfo Novak)
The project is authored and released under the nvk GitHub/X identity. Public biographical coverage identifies nvk as Rodolfo Novak, founder of Coinkite and a Bitcoin hardware/security entrepreneur; the LLM Wiki repository itself attributes the software copyright to nvk.
Founding Story
nvk launched LLM Wiki after Andrej Karpathy's April 2026 LLM-wiki concept, describing the implementation as something that 'had to be done' and 'surprisingly viable.' The initial vision was a Claude Code plugin that could initialize a topic wiki, conduct multi-agent research, and turn sources into durable, cross-referenced knowledge and deliverables rather than leaving the work in a chat transcript.
Target Markets
- Deep research and thesis testing across academic, technical, applied, news, and contrarian sources
- Building durable personal or team knowledge bases for AI coding agents
- Turning research into implementation plans, project briefs, reports, and other deliverables
- Maintaining source provenance, freshness, inventories, datasets, and follow-up state
- Privacy-conscious capture, auditing, and reuse of agent sessions
- Local-first, read-only querying of large Markdown knowledge bases