/mind-maps-for-agents
Your AI agent builds the mind map. You watch it appear.
/why-a-map
Agents are good at structuring ideas. A long chat reply is a bad place to keep them.
MindMap gives your agent somewhere to put that structure. It builds the map over MCP, and you see it land on your canvas while it works: a root, its branches, each branch in its own color.
Then you take over. Edit any node from the keyboard, and your agent sees your changes the next time it reads the map.
Works with Claude Code, Claude Desktop, Cursor and any MCP client.
Already have an account? Sign in
Connect to my MindMap account so you can create and edit mind maps for me.
MCP server: https://mindmap.jpamorgan.com/mcp (Streamable HTTP, OAuth)
- Claude Code: run `claude mcp add --transport http mindmap https://mindmap.jpamorgan.com/mcp`, then run /mcp and sign in.
- Other clients: add a remote MCP server with that URL. It opens a browser so I can sign in.
Once connected, call list_mindmaps to confirm it works. When I ask you to map something, call create_mindmap with a nested tree (3–7 branches, short labels), then give me the link./how-it-works
How it works.
Your agent does the setup. You sign in and approve it once, and from then on you ask for maps in plain words.
You copy the prompt.
Paste it into Claude Code, Claude Desktop, Cursor or another MCP client.
Your agent connects.
It adds the MindMap server and opens your browser. You sign in and approve what it can do. It never sees your password.
The map appears.
Ask for a map. It lands on your canvas as the agent builds it, and every later edit shows up the moment it’s made.
RootThe topic you asked about.BranchEach branch gets its own color, and everything under it shares it.Agent editNew nodes from your agent flash blue as they land.
/tools
Tools.
What your agent can do with your maps. Each tool runs the same code as the web app, so your agent gets the same results and the same errors you would.
- 01
list_mindmapsList the user's mind maps, most recently updated first, with each map's url.
- 02
get_mindmapRead one mind map: its version, url, nested tree with node ids, and a Markdown outline where each node begins with its [n_id] tag.
- 03
create_mindmapCreate a mind map from a nested tree and return a receipt with its url and new node IDs; give the url to the user.
- 04
edit_mindmapApply a batch of operations to one map atomically: all succeed or none do.
- 05
delete_mindmapPermanently delete a whole map.
- 06
get_mindmap_activityList recent changes to one map, newest first: who made them (the user or a named agent) and what changed.
/agents
Agents.
Anything that speaks MCP over HTTP and signs in with OAuth. Paste the prompt and the agent sets itself up, or add the server by hand.
Claude Code
One command, then /mcp to sign in.
Set up Claude Codeclaude mcp add --transport http mindmap https://mindmap.jpamorgan.com/mcpClaude Desktop and claude.ai
Add a custom connector with the server URL.
Set up Claude Desktop and claude.aihttps://mindmap.jpamorgan.com/mcpCursor
Add the server to mcp.json.
Set up Cursorhttps://mindmap.jpamorgan.com/mcpAny MCP client
Streamable HTTP, with OAuth and dynamic client registration.
Set up another clienthttps://mindmap.jpamorgan.com/mcp