A lightweight wiki that puts AI agents and humans on equal ground.
Pages are plain markdown with SQLite FTS5 search, wikilinks, and backlinks.
One command to install a true native binary, set up HTTPS at mind-map.local,
and configure every MCP client on your machine. Zero manual setup.
curl -fsSL https://github.com/aniongithub/mind-map/releases/latest/download/install.sh | bash
Native Windows binary — no WSL required.
The idea of LLM-maintained knowledge bases is gaining traction — agents that build and curate a persistent wiki rather than re-discovering answers via RAG. Tools like Obsidian and Tolaria are built with the user in mind, not the agent. mind-map puts both on equal ground.
Existing tools require:
Give agents a wiki. It runs anywhere:
mind-map.local — no warningscurl | bash and you're doneOne protocol. One server. Two kinds of clients. Edit at https://mind-map.local — agents see the change instantly.
Agents use stdio MCP. Humans use the web UI at https://mind-map.local. Both read and write the same markdown files.
Browse, search, and edit your wiki from any browser. Metro-inspired, blazing-fast chromeless design.
All in one native executable.
Full-text search powered by SQLite FTS5. Ranked results with highlighted snippets.
[[page]] syntax rendered as clickable links. Backlinks shown on every page.
Click edit, change the markdown, save. The search index updates instantly.
It's a server, not an app. Run it wherever you need it.
Install with one command. The installer adds a desktop shortcut — click to launch your wiki and open the browser.
Single static binary. Run in Docker, Kubernetes, a cloud VM, or as a sidecar. Multiple agents share the wiki safely via SQLite page locking.
Agents map wiki prefixes to GitHub repo wikis with register_sync. Knowledge syncs automatically via git.
Everything an agent needs to use the wiki as persistent memory.