LLM Wiki

LLM Wiki is an open source toolkit that lets AI coding agents build and maintain source backed knowledge bases in plain Markdown. One command creates a topic wiki, sends out parallel research agents, ingests the sources worth keeping, compiles them into cross referenced articles and then produces deliverables such as reports, slide decks, study guides, playbooks and implementation plans on top. It is aimed at developers and researchers who already work inside agent harnesses and want those agents to accumulate durable, queryable knowledge instead of starting from scratch each session. It ships as a Claude Code plugin, an OpenAI Codex plugin, an OpenCode instruction file, a skill for the Pi agent, or a portable AGENTS.md file for any other agent, and the resulting wikis open in Obsidian. Research mode runs five to ten agents across academic, technical, applied, news and contrarian angles, and can keep going in rounds for a set time. Thesis mode starts from a claim and returns a verdict rather than a summary. Ingestion accepts URLs, files, PDFs, Git documentation repositories, MediaWiki dumps, message archives and Wayback snapshots, while raw sources stay unchanged. Other features include Project Knowledge Checkpoints that export related knowledge as a five file handoff with privacy checks, personal specialist review methods, idea to project incubation, topic archiving, dataset manifests, session digests and a compact read only query command. LLM Wiki is free and released under the MIT license, with source code and releases on GitHub. There are no paid plans; the only costs are those of the AI agent and model the user runs it with.