llmwiki

Your coding-agent sessions, compiled into a local wiki you can search, graph, and open as plain files.

Live demo · Karpathy's LLM Wiki pattern

llmwiki turns transcripts from Claude Code, Codex, Cursor, OpenClaw and others into redacted markdown, then builds a static site: session browser, topic graph, candidate review, MCP tools. Your data stays in a vault outside git. Nothing has to be running to read the site — open site/index.html.

What you get

• A searchable static site — sessions, projects, topics, graph, analytics. Dark mode, Cmd+K search, syntax highlighting. Open the files; there is no server.
• A knowledge layer — synth writes source summaries and harvests entity/concept candidates. You review those on /candidates.html and apply decisions with llmwiki candidates apply. That review is a real gate, not an optional extra.
• Exports agents can read — llms.txt, llms-full.txt, JSON-LD, sitemap, RSS, plus an MCP server.
• Your data never lands in git. Transcripts, wiki pages and the built site live in a vault directory you choose. This repository is code plus a public demo.

Page kinds and where each field comes from: docs/reference/page-kinds.md. Every screen: docs/reference/ui.md.

Install

Requires Python ≥ 3.12 (CI verifies 3.12 only; newer interpreters are installable but not matrix-tested). Runtime dependency is markdown only.

pip install llm-wiki-plus
# or: brew install AlexanderMakarov/tap/llmwiki
llmwiki init --vault ~/llmwiki-vault

Point a gitignored config.json at the vault so later commands need no --vault:

{
  "vault": { "default_path": "/home/USER/llmwiki-vault" }
}

From a clone, ./setup.sh installs the package and reports which agents it can see. The clone itself is not a vault — bare llmwiki init / sync / build against this repository are refused; name --vault or demo.

Agent commands (slash commands and skills) ship inside the package. After install:

llmwiki install-agent-kit --dest ~/.claude

That is what a Homebrew or pip user runs so /wiki-sync and friends work from their project, not from this repository. See docs/reference/cli.md.

The install also drops a small .llmwiki-agent-kit.json manifest in --dest, recording which files it wrote and their checksums. It is safe — and useful — to commit alongside .claude/commands/: it is what lets a later upgrade recognise its own files and remove ones it has retired, while leaving anything you wrote or edited alone.

The loop

sync / add  →  synth  →  review candidates  →  build  →  open site/index.html

llmwiki sync                         # new transcripts → vault/raw/sessions/
llmwiki add notes.md                 # optional: drop a document into raw/docs/
llmwiki synth                        # source pages + harvest candidates
# open vault/site/candidates.html, set decisions, paste the printed command
llmwiki build                        # vault/wiki/ → vault/site/

synth does not rebuild the site. build does. candidates apply rebuilds after a successful batch so /candidates.html matches the wiki (--no-rebuild skips that). Candidate review is the human gate: a row starts with no decision, Apply assembles llmwiki candidates apply --vault … --actions '…', and only that command writes the wiki.

One-shot: llmwiki all --graph-engine builtin runs the whole loop — sync → synth → build → graph → lint. Opt out of a stage with --no-sync, --no-synth, --skip-graph, or --skip-lint.

Agents

One row per agent. Supplies sessions = llmwiki sync can read its transcript store. Reads the wiki = slash commands / MCP after install-agent-kit (any MCP client can also call python3 -m llmwiki.mcp). Core adapters run on a default sync when their store exists; contrib adapters need --adapter <name> or adapters.<name>.enabled in config.

llmwiki adapters --wide
llmwiki sync --adapter openclaw

Configuration

Shipped defaults: examples/sessions_config.json. Personal overrides: gitignored config.json at the clone root, merged on top. Full keys: docs/configuration.md.

Skip projects or dates with a .llmwikiignore next to the vault (one path or glob per line).

Documentation

This repository's CLAUDE.md and AGENTS.md are for people changing llmwiki itself. Vault workflows for users live in the installed agent kit.

Acknowledgements

• Andrej Karpathy — LLM Wiki idea
• Pratiyush — upstream this work extends

License

MIT © Alexander Makarov; based on upstream Pratiyush/llm-wiki