> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.livingbrain.com/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.livingbrain.com/_mcp/server.

# Core concepts

Living Brain is a compiler. You push raw material in; it produces a maintained set of pages
and the relationships between them.

```
capture ──▶ ingest source ──▶ compilation ──▶ page(s) ──▶ connections
                                                  │
                                                  └──▶ change event
```

## Brain

A **brain** is an isolated knowledge vault. Every capture, page, and connection belongs to
exactly one brain, and every brain is owned by one subject (see [Authentication](/docs/authentication)).
Most products create one brain per end user, though nothing stops you from giving a single
user several — one per project, say.

## Capture

A **capture** is raw material you push in. Six kinds:

| `kind`        | What it's for                             |
| ------------- | ----------------------------------------- |
| `note`        | Free text the user wrote                  |
| `url`         | A link to fetch and read                  |
| `file`        | An uploaded PDF, DOCX, or PPTX            |
| `transcript`  | Meeting or call transcript                |
| `chat_turn`   | A single conversational turn — see below  |
| `integration` | Material arriving from a connected system |

Capture is **asynchronous**. The call returns an ingest source, not a page.

## Ingest source

An **ingest source** is the record of one capture moving through the pipeline. Its `status`
walks through `pending` → `compiling` → `completed`, or lands on `failed`. Failed sources can
be retried without re-uploading the original material.

## Page

A **page** is the compiled unit of knowledge — the thing the brain actually maintains. Pages
are addressed by `slug` within a brain, carry a `title`, `summary`, and `body`, and are
versioned: every edit or recompile produces a new version.

Each page has a `pageType`:

`source` · `topic` · `entity` · `project` · `decision` · `brief`

and a `status`:

| Status       | Meaning                                        |
| ------------ | ---------------------------------------------- |
| `active`     | Current and trusted                            |
| `stale`      | Hasn't been referenced or refreshed recently   |
| `contested`  | New material contradicts what the page asserts |
| `deprecated` | Superseded, hidden from listings by default    |
| `archived`   | Retired by the user                            |

> **Info**
>
> `contested` is the interesting one. When a new capture disagrees with an existing page,
> Living Brain flags it rather than silently overwriting — then you decide, via
> [`resolve`](/api-reference).

## Bucket

Buckets are the top-level grouping: `inbox`, `notes`, `ideas`, `projects`. Every page sits in
exactly one. `bucketSource` records whether the brain routed the page there (`auto`) or a
human moved it (`user`) — useful when you want to avoid re-routing something a user placed
deliberately.

## Connection

A **connection** is a discovered relationship between two pages, with a `strength` of `weak`,
`medium`, or `strong`, and a `source` of `auto` or `manual`. Connections are what make the
brain more than a folder of documents — they're the non-obvious links across conversations,
documents, and time.

Users can dismiss a connection without deleting it, which keeps it out of unacknowledged views.

## Change event

Every meaningful mutation emits a **change event** with an `action`:

`created` · `updated` · `renamed` · `merged` · `split` · `archived` · `restored` · `recovered` · `routed`

The change feed is the primary way to keep an external system in sync. See
[Change feed and webhooks](/docs/changes).

## Brain profile

The **profile** holds standing directives that shape how the brain interprets new captures —
domain vocabulary, what the user cares about, how aggressively to merge. Profile changes apply
to subsequent captures, not retroactively.