Vinc

What Vinc is

A Vinc graph is made of six things: node kinds, relations, documents and chunks, topics, time and spaces.

Press a node to see its kind and role.

Node kinds

Each kind of node is handled differently. There are five kinds you create yourself.

ConceptAn idea you will point to again and again. Search results gather here, and other nodes relate to it.
DecisionWhat was chosen, and why.
RecordSomething observed or measured, with the period it held for.
EpisodeSomething that happened on a particular day.
BridgeA node that joins separate topics.

For example, “we connected Claude to Vinc so we share one graph” becomes three nodes: Vinc is a Concept, the choice to share one graph is a Decision, and the day you first connected Claude is an Episode. Put all of it in one node and “why did we pick this?” gets hard to answer later.

Record or Episode: “the build takes eleven minutes” can stop being true, so it is a Record. “the build failed on Tuesday” already happened, so it is an Episode.

A kind that fits none of the five is stored as an Entity, and its new name stays a proposal until a person approves it. Entities can relate to each other under any relation name you choose. An Entity does not keep props or domain, so use a Concept or a Record for anything that has to carry values.

You normally do not create Domain nodes. Vinc creates them from the slashes in a topic.

Relations

A relation has a name and a direction, and reversing it changes what it says. There are six relations you create yourself.

RelationMeaningFrom → to
relates_toLoose associationConcept → Concept
prerequisite_ofHas to come firstConcept, Decision → Concept, Decision
derives_fromWhere it came fromConcept, Decision → Concept, Decision
inherits_fromWhat it inheritsConcept, Decision → Concept, Decision
aboutWhat it is aboutEpisode, Decision → Concept, Decision
connects_viaWhat a Bridge joinsBridge → Concept, Decision

For example, prerequisite_of from mcp to ai-assistant says “MCP has to be connected before an AI assistant can use the graph.” Reverse it and it says the opposite.

From a Record, the first four relations (relates_to, prerequisite_of, derives_from, inherits_from) can point to a Concept or a Decision.

A pair that is not allowed is not stored, and a warning comes back. The system draws MENTIONS itself, and ATTACHES_TO and TYPED are how relations from a Record or between Entities are stored. You never write any of the three.

Documents and chunks

A document is stored in the graph exactly as written, so the text stays even if you delete the file you imported. It is split into chunks for search, and joining the chunks gives back the original exactly. On read, the offsets and a hash are checked, and a document that passes both is marked verified.

When a document comes in, Vinc finds the concepts and decisions the text names and links them with MENTIONS. The links follow fixed rules with no model involved, so the same document always gives the same result.

A chunk is not a summary or an excerpt. It is a piece of the original.

Topics

A topic is a string divided by slashes. Write ai/mcp and mcp sits under ai, with no relation to create. Filtering by a topic includes its subtopics.

Case matters. AI/MCP and ai/mcp are two different topics.

A topic is not a permission. Splitting topics hides nothing from anyone; the space decides who can see a node.

Time

Vinc records time in three ways.

content timeWhen the thing actually happened. Concepts, records and documents carry it; an episode carries its date.
graph timeWhen it was recorded in the graph. The system stamps it.
valid timeThe period a record or a decision holds for. Start and end are written separately.

For example, if you record this week a decision made last spring, its content time is last spring and its graph time is this week.

A record or decision that no longer holds gets an end date. It drops out of the default search but is not deleted, so you can still see what held at the time.

Spaces

A space is the scope a graph belongs to. Your personal graph is yours alone; a team graph is shared with its members. Every read and write happens inside one space.

In a team graph each member has one role: viewer, editor or admin. Team graphs are available on the web and in the desktop app's account mode; local mode has none.

If you do not name a space, tools write to your personal graph. Name the space whenever you mean a team graph. Nodes in your personal graph stay there until you copy them with Tools > Share.

Learn more

For the full story, read the Book.