Axiom Wiki.
Axiom Wiki is a CLI-first personal knowledge system that ingests files, URLs, codebases, and notes, extracts structure, compiles them into cross-linked markdown pages, and exposes the result to humans and AI agents through search, graph tools, and MCP.
The Challenge
Most personal knowledge systems store notes, but they do not actively compile source material into a maintained, queryable knowledge structure. RAG can answer questions, but it often re-derives context from raw files instead of preserving a durable map of what is known.
- Project and research knowledge getting trapped in scattered documents
- Agents repeatedly rediscovering the same context
- Raw RAG answers disappearing instead of becoming maintained documentation
- Manual wiki upkeep becoming too expensive as files change
- Users needing local-first, markdown-first control over their knowledge base
The Solution
I built Axiom Wiki as a knowledge compiler: sources become plain markdown pages with links, indexes, search, graph structure, agent instructions, and MCP tools so both people and agents can reuse the same maintained knowledge base.
- Published as an installable CLI with `axiom-wiki` and `axwiki` commands
- Supports document ingestion, autowiki generation, sync, watch mode, web clipping, search, graph, and MCP serving
- Works with multiple cloud and local model providers instead of one fixed LLM
- Keeps output Obsidian-compatible through plain markdown pages and frontmatter
Technical Depth
Document Ingestion
Ingests markdown, PDFs, DOCX, images, HTML, and clipped web pages, then turns raw sources into durable wiki pages instead of leaving them as disconnected files.
Auto-Wiki Mode
Lets an agent explore a folder or codebase, decide what pages are missing, and build a useful wiki in batches without requiring a hand-written outline.
Incremental Compilation
Tracks source changes and avoids reprocessing unchanged files, so the wiki can grow without turning every update into a full rebuild.
Self-Maintaining Sync
Detects stale pages, updates renamed or changed areas, and keeps the generated knowledge base aligned as the underlying project evolves.
Hybrid Search & Graph
Combines keyword and semantic search with page graph visualization so the wiki can be browsed, queried, and inspected from multiple angles.
MCP for Agents
Exposes wiki querying, source inspection, and maintenance tools to Claude Code, Cursor, and other MCP-compatible clients.
Provider Flexibility
Supports Gemini, OpenAI, Anthropic, OpenRouter, DeepSeek, Groq, Mistral, and local Ollama workflows so users are not locked into one model stack.
Agent Setup Templates
Generates instructions for coding agents so they know how to use the local wiki as project memory instead of repeatedly rediscovering context.
Health & Cost Visibility
Surfaces wiki status, staleness, confidence, semantic index health, token usage, and operation cost in the CLI and local files.
Technology Stack
Highly specialized modern technology stacks utilized to achieve system goals.