introduction

---

title: Introduction type: doc

Introduction

mind-map is a wiki engine designed from the ground up to be edited by both humans and AI agents, sharing the same files, the same index, and the same semantics.

The problem

Modern AI agents need persistent memory: notes, decisions, prior conversations, project state, design rationale. The popular options all force a tradeoff:

Tool	Agent-friendly?	Human-friendly?	Self-hosted?
ChatGPT memory	yes	no	no
Vector DBs (Pinecone, Weaviate)	yes	no	sometimes
[[comparisons/notion	Notion]]	via clunky API	yes
[[comparisons/obsidian	Obsidian]]	local files only	yes
[[comparisons/logseq	Logseq]]	local files, no API	yes
Tolaria	yes	yes	yes (heavy)
mind-map	yes (MCP)	yes (Web UI)	yes (38 MB)

How mind-map approaches it

Three foundational decisions:

1. Storage is the filesystem. Pages are just markdown files. See concepts/pages. You can git clone your wiki. You can grep it. You can edit it in vim while the server's running.
2. The agent interface is architecture/mcp-server. Any agent that speaks Model Context Protocol gets read/write/search/move tools for free.
3. The human interface is a architecture/web-ui that renders the same files, the same concepts/wikilinks, and the same concepts/backlinks the agent sees.

What you get

• A 38 MB single binary (design/lightweight)
• concepts/search across the whole wiki
• A architecture/web-ui of pages and their links
• concepts/wikilinks you can refactor with agents/mcp-tools without breaking history
• concepts/frontmatter for typed metadata

Next

• guides/quickstart — get it running in 30 seconds
• guides/install — the one-click installer in detail
• guides/platforms — what runs where
• guides/out-of-the-box — the defaults
• guides/service — run as a system service
• concepts/index — read about the primitives
• design/rationale — read the longer-form design notes