Content
Content is everything you write in the hub. Docs, notes, research papers, narratives and discussion posts are all documents written in markdown, and every one of them is kept the same way: with a history of versions, searchable, and filed in a place in Storage that says what kind of document it is. What differs is the kind.
Each kind has a feature of its own, and discussion posts belong to Community. Beneath all of them are the things that organise a document, such as categories and tags, and the files and links you attach to one.
Topics
- Markdown documents: what every document has, and how it is kept and searched.
- Docs: your documents.
- Notes: your notes.
- Research: papers you write and can publish on the research site.
- Narratives: your narratives.
- Block documents: a second kind of document, edited together in blocks.
- Categories, tags and keywords: the three ways a document is filed.
- Attachments and URLs: files and links kept with your content.
- Personal data: your profile: jobs, education, contacts and the rest.
Everything you write on the site, a program can read and write over the API: documents of every kind, their versions, the categories and tags they are filed under, the files and links attached to them, and your personal data.
Markdown documents are the content routes for markdown, with list, create, read, update and delete. Search across everything you can read is one call, and the recent-activity feed is another.
Reactions, poll votes and bookmarks are content routes too, and they attach to any markdown document.
Personal data is a set of plain resources, one for each list, with the same list, create, read, update and delete as any other table.
Over MCP, markdown documents are on the agent allowlist, so an agent with a scoped token reads and writes them. The agent can also search across your data, read the contents of a category, search your notes, and read the keywords the AI processing node gave a document. The profile tool group reads your personal data: your profile, jobs, education, locations, contacts, relationships, tags and dates.
The full list of routes, with each one's parameters, body and responses, is the REST reference on the help site. The tools an agent gets are listed in the MCP reference.
One row shape, many shelves. One row shape serves every kind: a marker and a bucket tell them apart. The marker and the bucket say what shelf a document is on; the shape never changes. That is why search, versions, reactions, embedding and recall are written once.
Categories are a web, and the database keeps it from looping. A document can be in more than one category and a category can have more than one parent, and a database trigger makes a cycle impossible no matter which client writes the link.
Each piece carries its own staleness. A piece of a document records whether its embedding still matches its text, so a re-embed touches only what changed.
Research is public by design. It is the one workspace segment search engines may index.