Capture a new idea or concept into the persistent mindmap. ONLY use this when the user themselves has expressed an idea, question, goal, task, topic, or interest — including casual searches ('how to cook biryani'), projects they mention, or anything they bring up. NEVER invent, generate, or seed example nodes. NEVER add content Claude thought of. Before adding, check if the same idea already exists (use search_ideas) to avoid duplicates. Attach it to an existing node with parentId to build hierarchical structure. Returns the new node with its ID (needed for future parentId references).

Edit the text or tags of an existing node. Use this when an idea has been refined, clarified, or its scope has changed. Does not change parent/child relationships — use add_idea + delete_node to restructure.

Remove a node from the mindmap. Child nodes are orphaned (kept in the map but disconnected from the tree). Use search_ideas first if you are unsure of the nodeId.

Search the mindmap for nodes matching a query string. Scores nodes by exact match > substring > per-token matches in text and tags. Use this to find a node ID before updating/deleting, or to surface related ideas.

Retrieve the mindmap as a nested tree. With no arguments, returns the full tree from the root plus any orphaned nodes. Pass nodeId to get just that subtree (the node and all its descendants).

Export the full mindmap as raw JSON — the exact contents of mindmap.json. Use this to save a backup, import into the visualizer at the project website, or inspect the raw data structure.

Export the full mindmap as a Markdown nested list, ready to copy into a document or note. Tags are shown inline. Orphaned nodes are listed in a separate section.

Export the full mindmap as a Mermaid flowchart diagram. Paste the output into any Markdown renderer that supports Mermaid (GitHub, Notion, Obsidian) to get a visual graph.

Export the full mindmap as OPML (Outline Processor Markup Language). OPML is supported by dedicated mindmap apps such as MindNode, OmniOutliner, and XMind for import.

Generate a self-contained interactive HTML mindmap and save it to ~/.mindkeeper/mindmap-export.html. After calling this tool, tell the user to open the file in their browser — no server required. The diagram is draggable, zoomable, and uses a side-branch layout with curved connections. The toolbar inside the HTML has buttons to download the diagram as PNG or SVG.

Read a conversations.json file exported from Claude.ai and return a structured list of conversations. The user must first download their data at claude.ai → Settings → Account → Export Data, unzip the archive, and provide the path to conversations.json. After calling this tool, YOU (the AI) must analyze the returned conversation list and: 1) identify major topic clusters across the conversations, 2) call add_idea to create a root topic node for each cluster, 3) call add_idea to add notable sub-topics or recurring themes as children, 4) skip trivial, one-off, or very short conversations. Group by theme, not by conversation title. Aim for a clean, meaningful mindmap hierarchy. When done, call export_html so the user can visualize the result.

Sync the mindmap with GitHub Gist cloud storage. Requires ~/.mindkeeper/config.json with cloud credentials (see README). Use direction='push' to upload local state, 'pull' to download and overwrite local. First push auto-creates a private Gist and saves the gistId back to config.

Check whether cloud sync is configured and show the current config (token is masked). Returns the provider, gistId (if any), and setup instructions if not configured.