Your mind maps, as tools for AI agents

Requirements

• A miMind AI subscription, signed in to your miMind account.
• The miMind desktop or mobile app (the MCP dialog lives there).

Steps

1. Open the main menu → Settings → AI Settings and pick MCP Server… (available from both the map list and an open map).
2. Enable Sync my maps to the MCP server. From now on, every map you save publishes its outline so agents can read it.
3. Press Generate New Token and copy the token — it is shown exactly once (miMind stores only a hash). You can keep up to 5 active tokens and revoke any of them at any time.
4. Copy the Server URL from the same dialog and configure your MCP client with both (see below).

Claude Code (CLI)

claude mcp add --transport http mimind \
  https://us-central1-mimind-ebda5.cloudfunctions.net/mcp \
  --header "Authorization: Bearer mmcp_YOUR_TOKEN"

Claude Desktop / other MCP clients

Add a custom connector / remote MCP server with the same URL and an Authorization: Bearer mmcp_… header. The server uses the MCP Streamable HTTP transport (stateless), which every current MCP client supports.

Then just ask

List my mind maps What's in my “Product Roadmap” map? Create a mind map about container gardening for beginners

list_maps read

Lists your synced maps: id, title, node count, and when each was last updated.

Argument	Type	Description
— none —

Which mind maps do I have?

{ "jsonrpc": "2.0", "id": 1, "method": "tools/call",
  "params": { "name": "list_maps", "arguments": {} } }

Returns one line per map, e.g. - map_id: a1b2… | "Product Roadmap" | 42 nodes | updated 2026-07-24T…

read_map read

Returns a map's full outline as indented text — one node per line, children indented under their parents.

Argument	Type	Description
map_id	string, required	The id from list_maps.

Read my Product Roadmap map and summarize the Q3 branch

{ "jsonrpc": "2.0", "id": 2, "method": "tools/call",
  "params": { "name": "read_map", "arguments": { "map_id": "a1b2c3" } } }

search_maps read

Case-insensitive text search across all synced maps (titles and node text). Returns each matching map with the matching lines.

Argument	Type	Description
query	string, required	Text to search for.

Search my maps for anything about budget

{ "jsonrpc": "2.0", "id": 3, "method": "tools/call",
  "params": { "name": "search_maps", "arguments": { "query": "budget" } } }

create_map write

Creates a new miMind map from an indented text outline. The map is queued and built by your miMind app on its next sync (app start or map-list refresh) — it then appears in your map list like any other map.

Argument	Type	Description
title	string, required	The map title (becomes the central topic).
outline	string, required	Indented outline text — one node per line, children indented with 2 spaces or tabs; -/* bullets are accepted.

Create a mind map called “Trip to Japan” planning a two-week itinerary

{ "jsonrpc": "2.0", "id": 4, "method": "tools/call",
  "params": { "name": "create_map", "arguments": {
    "title": "Trip to Japan",
    "outline": "Trip to Japan\n  Week 1 - Tokyo\n    Shinjuku\n    Day trip to Nikko\n  Week 2 - Kansai\n    Kyoto temples\n    Osaka food tour\n  Budget\n  Packing list"
  } } }

add_nodes write

Queues new nodes for an existing map. They are applied the next time you open that map in miMind — with normal undo history, so you can always revert an agent's additions.

Argument	Type	Description
map_id	string, required	The id from list_maps.
parent_title	string, optional	Title of the existing node to attach under. Defaults to the map's root.
outline	string, required	Indented outline of the new nodes.

Add three risk items under “Risks” in my Product Roadmap map

{ "jsonrpc": "2.0", "id": 5, "method": "tools/call",
  "params": { "name": "add_nodes", "arguments": {
    "map_id": "a1b2c3",
    "parent_title": "Risks",
    "outline": "Supply chain delays\nKey-person dependency\nCurrency exposure"
  } } }

Agents never modify your data directly. create_map and add_nodes place a command in a per-account queue; your miMind app builds the map or applies the nodes, using the same engine as the built-in AI assistant:

• create_map — materialized at app start / map-list refresh; the new map lands in your active folder (and uploads if the folder is cloud-synced).
• add_nodes — applied when you next open the target map, with undo history.
• Reads (list_maps, read_map, search_maps) reflect the outline published on each save — a just-created map becomes readable after miMind builds and saves it.

• Transport: MCP Streamable HTTP, stateless. POST JSON-RPC 2.0 messages to the server URL; responses are plain JSON (no SSE stream, no sessions). Batch requests are not supported.
• Methods: initialize, notifications/initialized, ping, tools/list, tools/call. Protocol revisions 2025-06-18, 2025-03-26 and 2024-11-05 are accepted.
• Auth: Authorization: Bearer mmcp_… on every request.

curl -X POST https://us-central1-mimind-ebda5.cloudfunctions.net/mcp \
  -H "Authorization: Bearer mmcp_YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Status	Meaning
401	Missing, invalid or revoked token — generate one in miMind's MCP Server dialog.
403	No active miMind AI subscription, or AI access disabled for the account.
503	The MCP service is temporarily disabled.
Tool errors	Returned as MCP tool results with isError: true and a human-readable message.

• Tokens are hashed — miMind stores only a SHA-256 hash; the full token is shown once, to you. Revoke any token instantly from the MCP Server dialog.
• Sync is opt-in — nothing is published until you enable it, and only map outlines (titles and node text) are synced, never attachments or files.
• Scoped to you — a token only ever reaches the maps of the account that created it.
• Undoable — every agent change is applied by miMind through its normal editing pipeline and can be undone.
• External agents run under their own terms — any MCP client you connect (Claude Desktop, Claude Code, or another) processes what you send it under its own provider’s privacy policy.

For how miMind handles your data across accounts, cloud sync and AI, see the Privacy Policy (and its AI & Anthropic section).