Explore

The tree that teaches you. Queue topics. Build curricula. Study through conversation. Track mastery. Detect gaps. The AI guides you through the subject, asks questions, evaluates understanding, and adapts to your learning style. Integrates with the learn extension for URL content fetching. Mastery...

The tree learns what you like. Not from settings. From watching. Signals accumulate on nodes as users interact. AI-generated content that gets kept is positive. Content that gets edited is mild negative. Content that gets deleted is strong negative. Navigation frequency is implicit preference....

When a tree has been alive long enough, it accumulates wisdom that is not structure and is not content. It is meta-knowledge. Evolution discovered that branches with 3 children complete 4x more than branches with 10. Purpose learned that the thesis drifts when the user adds too many top-level...

How people work together on trees. Trees start with one owner. Team adds the ability to invite contributors, transfer ownership, remove users, and retire trees. The invite system is atomic: pending invites transition to accepted or declined in a single findOneAndUpdate to prevent double-processing....

Follow one thread through the entire tree. Every node it touched, in order. Not broad search like scout. Not downward exploration like explore. One concept, every note that references it across the whole tree, chronologically. Where did it start? How did it evolve at each stop? What's the current...

Nodes can trade value with each other. A transaction is a two-sided exchange: node A sends values to node B, and node B sends values back to node A. Or one side is OUTSIDE, representing an external source like a Solana wallet. The extension handles the full lifecycle: proposal, approval, execution,...

A tree outgrows its usefulness. Too many nodes. Too much metadata. The AI context window cannot hold the summary. The circuit breaker is approaching threshold. The tree needs to shrink without dying. tree-compress walks from leaves to root, compressing content at each level, carrying the essential...

Position determines reality. Every message that enters a tree passes through this orchestrator. It routes to the right extension or to tree:converse. The routing index maps extensions to positions in the tree. If the current node or a nearby node has a mode override (modes.respond), and the...

The first operating system built on the seed. Four bundles, twenty base extensions, nine standalone. Install once and the tree thinks, navigates, structures, edits, reads, writes, responds, cascades, compresses, detects contradictions, acts autonomously, breathes, remembers every age, and connects...

The reference implementation of how the AI thinks inside a tree. Eleven modes, thirty-plus MCP tools, and a navigation hook that keeps the frontend synchronized with every operation the AI performs. This is the foundation that every tree conversation builds on. Converse is the default mode for...

The nervous system. Eight extensions that make cascade real. Without this bundle, cascade is just a kernel hook that fires and nothing listens. The kernel provides the primitive: when content is written at a cascade-enabled node, fire onCascade. That's it. The kernel doesn't propagate. It doesn't...

The rain. Eleven extensions that open the clouds. Without this bundle, the only way data enters your land is through the CLI, the web interface, or the API. Your trees grow from what you manually plant. With this bundle, external channels become input sources. Discord messages become tree...

Self-awareness. Fourteen extensions that teach the tree to know itself. Tree-compress monitors density and consolidates structure. Contradiction detects conflicting truths across branches. Inverse-tree builds a model of the user from interaction patterns. Evolution tracks which organizational...

Hygiene. Five extensions that keep the tree clean and organized. Prune sheds dead weight. Identifies truly dormant nodes, absorbs their essence into the parent, trims the branch. The tree shed a leaf because autumn came. Reversible. Reroot reorganizes the living tree. Builds a semantic...

A tree with 500 nodes cannot fit in an AI context window. The AI needs to know what the tree contains without reading every note on every node. Understanding solves this by building a compressed encoding of the entire tree from the bottom up. Leaves first, then their parents, then the parents of...

Cross-tree queries scoped to a single user. The kernel stores notes, contributions, and chats per node, but users need to see their own activity across all trees in one place. This extension provides three endpoints: /user/:userId/notes returns all notes the user has written across every tree they...

The access control layer between users and features. Every user has a tier stored in metadata.tiers.plan (defaults to 'basic' if unset). Extensions declare which tiers unlock their features by calling registerFeature(featureName, allowedTiers) during init. The hasAccess(userId, feature) export...

Not everything in a tree is text. Some things are numbers. Revenue, hours logged, calories, word count, completion percentage, test coverage, budget remaining. Values attaches named numeric key-value pairs to any node. Goals attaches target numbers to those same keys. Together they answer two...

The full picture at any position. Combines perspective, codebook stats, memory, gaps, flow, and evolution into one view. What is flowing through this node right now? Perspective shows what it drinks. Memory shows who it has talked to. Gaps show what it is missing. Flow shows recent signals....