Agent Wiki CLI
The awiki CLI keeps agent memory in a markdown vault: search before the web and save whatever is worth keeping.
- 56 ★ on the repository
- Last commit
Extended access
Why this level
- Hooks edit agent settings, and ingest downloads pages and stores their text in the vault.
- The shared server serves memory over HTTP with tokens, so ordinary service hardening applies.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/agent-wikiDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Hooks edit agent settings, and ingest downloads pages and stores their text in the vault.
- The shared server serves memory over HTTP with tokens, so ordinary service hardening applies.
This is third-party code. Review the repository files before installing.
What it does
Agent Wiki turns a folder of markdown files into shared agent memory. The agent searches it before the web and, after working, saves what is worth keeping. Inside there are only files with a YAML header and `[[wikilinks]]`, no database, and the vault can be kept in Obsidian. The single entry point is the `awiki` CLI. `awiki ingest` copies a file or fetches a page into an immutable `raw/` archive and writes a sha256 checksum. `awiki search` runs a multi-word search, `awiki show` prints a page whole. Claude Code, pi and OpenCode have hooks, and `awiki guide` prints a block for `CLAUDE.md` or `AGENTS.md` that teaches the agent to search the wiki first.
Who it is for. People who work with several agents and projects and do not want solved problems lost when a session closes.
Good fit when
- The agent solved a stubborn build error and the write-up must survive into the next sessions.
- An article or PDF on a topic goes into an archive with integrity checks and links from pages.
- One vault server serves several machines and projects.
- Notes should be read and edited in Obsidian while the agent writes into the same files.
Not a fit when
- A vector database with semantic search is needed: search here is word-based, built on ripgrep.
- Every session must start from a blank slate, with no accumulated memory.
- The GPL-3.0 license conflicts with project policy.
Example request
Search the wiki for a solution to this error, and if it is new, save the write-up as a separate page with tags.Limitations
The CLI needs Python 3.10 or newer; the package is `agent-wiki-kb` and installs the `awiki` and `agent-wiki` commands. Session hooks write into agent settings: `awiki hook install --agent claude` edits `~/.claude/settings.json`. The network server `awiki serve` shares the vault over HTTP with tokens and reader, writer and admin roles, and must be protected like any service. Fetched pages and PDFs are saved into the vault whole, so private material is best kept out. The license is GPL-3.0 or later.
How to disable. Run `awiki hook uninstall`: it removes only what awiki wrote. Then delete the package and the vault folder if the memory is no longer needed.
FAQ
How do I install awiki?
Run `pip install agent-wiki-kb`; `uv tool install` and `pipx install` also work. Then create a vault with `awiki init <path>`.
How do I connect Claude Code?
The `awiki hook install --agent claude` command adds a `SessionStart` hook to `~/.claude/settings.json`. For pi and OpenCode use `pi` or `opencode` instead of `claude`.
How does the agent know to use the wiki?
Run `awiki guide` and tell the agent to paste the printed block into `CLAUDE.md` or `AGENTS.md`. The block teaches it to search first, read a page whole and save what helps.
Can the vault be shared across machines?
Yes: `awiki serve` publishes the vault over HTTP and other machines run the same `awiki`. Connect with `awiki init --remote <url> --token <secret>`.
Is a database required?
No. The vault is markdown files; read them with `grep`, open them in Obsidian or process them with Python scripts.
Discussion
No questions or reviews yet.
Related
CLI toolsLocal memory for AI agents in plain Markdown files: it keeps only what you approve and serves it through MCP
The lwc CLI in Rust: persistent agent memory as a source-grounded wiki on SQLite, for Claude Code, Codex, Cursor and more
agentcairn keeps Claude Code, Codex and Cursor memory in Markdown files for Obsidian and hands an agent only what it needs
Long-term memory for AI agents: Markdown files as the source of truth, a SQLite index as a cache, shared by Claude Code, Codex and others
Persistent agent memory in a folder of linked markdown notes, so Claude Code remembers a project across sessions
ai-memory retains session knowledge in a Markdown wiki, exposes MCP retrieval and passes context between agents and machines.
README in short
The vault format is close to the Google Open Knowledge Format. `index.md` and `log.md` are generated automatically, and no taxonomy is prescribed. The idea is inspired by a Karpathy note on wikis for language models.