Browse / LLM wiki / Second brain / agent-wiki
agent-wiki
xccElephant
A persistent, agent-maintained wiki for raw sources, notes, and research.
Open xccElephant/agent-wiki → Checked against the four rules.
agentclaude-codecodexknowledge-basellmmarkdownobsidianpersonal-knowledge-managementwiki
Works with any agent
ships AGENTS.md, the cross-tool instruction file.
In its own words
Works with Cursor, Claude Code, and Codex-style agents.
A persistent, agent-maintained wiki for raw sources, notes, and research.
Drop source material into raw/. Your agent reads it, extracts knowledge, cross-links ideas, and maintains a Markdown wiki that compounds over time.
Obsidian is the best viewer for the result, but the knowledge stays in plain Markdown and works with Cursor, Claude Code, and Codex-style agents.
Why This Exists
Most note systems still make you do the bookkeeping yourself:
- read the source
- summarize it by hand
- create concept pages
- update related notes
- remember what changed
agent-wiki turns that maintenance loop into an agent workflow. You keep the raw sources. The agent maintains the derived wiki.
Quick Start
- Clone this repo into a new vault or an empty folder:
- Bootstrap the starter files:
- Add files to
raw/articles/, then open the folder in Obsidian and your coding agent.
Ask your agent to ingest a source, update related wiki pages, or answer questions from the compiled wiki.
What You Get
raw/keeps immutable source snapshotswiki/stores derived knowledge pageswiki/source-notes/captures one note per sourcewiki/entities/,wiki/concepts/, andwiki/synthesis/separate durable knowledge typeslogs/log.mdrecords ingest and maintenance historyAGENTS.md,CLAUDE.md, and.cursor/rules/define the workflow for different agents
The repository's own description, then 40% of its README. Read the rest on GitHub ↗
Read from the repository's git tree on 2026-08-17.