Tana MCP. Turn unstructured text into structured knowledge graphs.

Listen up: The Tana MCP Server lets your AI agent treat your notes like a living database, not just some pile of text you copy-pasted from Reddit. It takes raw conversations and turns them straight into structured knowledge graphs that actually work for you.

You're done with copy-pasting; you prompt for structure. Your agent doesn't just summarize what you said—it builds the framework around it. You can make your AI client define entire classification schemas (Supertags) using define_supertag, making sure every thought you capture fits a strict organizational mold. When you write something down, your agent uses that schema to build out complex records; for instance, it'll run add_node_with_fields when you need an advanced note type populated with specific JSON data fields.

Need to organize your thoughts into a deep outline? The agent handles the nesting. It executes add_node_with_children, which builds a parent node and automatically drops several blank child nodes underneath it, letting you build out complex, nested conceptual maps right away. If you just need a basic idea captured somewhere—like dropping a quick thought into your main Inbox or Library—the agent uses add_node to get that simple text node down fast.

It's not just about making notes; it's about making them actionable. When the meeting wraps up, forget manually assigning tasks and dates. The agent converts raw minutes into functional components: you can use add_checkbox_task for a simple to-do item that lives in your Tana inbox until you check it off.

It handles time markers too; running add_date_node drops a node formatted specifically for the date (YYYY-MM-DD) right where you need it.

Linking ideas together is where this thing really shines. Instead of copy-pasting chunks of text, your agent uses add_node_reference, creating explicit links to existing notes so all concepts are traceable back to their source material. You can also quickly drop an unclassified thought anywhere using add_to_inbox or create a specialized bookmark for external sites with add_url_bookmark.

If you're capturing information that needs a specific label, like designating it as a meeting note, the agent uses add_tagged_node to create the node and slap the designated Supertag right on it.

This system bypasses manual entry entirely. The agent handles all the classification, linking, and structuring work for you. It builds multi-level hierarchies by creating parent nodes, then ensures that every piece of information is properly categorized and linked within your graph, keeping your focus local but making sure nothing gets lost in the shuffle.