Capacities

Promptfree, for everyone, and the only version there is

Copy the prompt for Claude CodeCopy the prompt for CodexCopy the prompt for Cursor

Copy

Build me the object-based notes system that replaces Capacities.

The idea worth copying: **organise by what a thing is, not by where you put it.** A folder forces one location; a type lets a note about a meeting be a meeting, linked to the people and the project, findable from all three.

STACK
- Node 20+ with Fastify, server-rendered HTML with a little vanilla JS
- SQLite through better-sqlite3, WAL mode, FTS5 across everything
- Markdown on disk as the source of truth, the database as an index
- Caddy in front

THE DATA MODEL
- objects: id, type_id, title, body_md, created_at, updated_at, path
- types: id, name, icon, fields_json, template_md
- values: object_id, field_key, value_text, value_object_id — a field pointing at another object is what builds the graph
- links: from_object_id, to_object_id, context — with the surrounding sentence
- daily: date, object_id

TYPES WORTH HAVING BY DEFAULT
- Person, meeting, project, book, article, idea, place. Seven types cover most of what anybody records
- Each with a few fields and a template. **Keep the field count small** — a type with fifteen fields is a form, and nobody fills in a form to write a note
- A type can be added or edited at any time, and existing objects gain the new field empty, with no migration

THE DAILY NOTE AS THE ENTRY POINT
- Everything starts in today's note. Mention a person and it links; write about a meeting and turn that line into a meeting object with one key
- **Converting a line into a typed object without leaving the flow of writing is the whole ergonomic argument.** If it takes a dialogue, people write into the daily note forever and nothing gets structured

VIEWS PER TYPE
- A list of every object of a type, with its fields as columns, sortable and filterable
- On a person: every meeting they were in, every note mentioning them, gathered automatically from the links
- **That automatic gathering is what replaces filing.** Nothing is put anywhere; everything is found by relationship

SEARCH AND LINKS
- FTS5 across titles and bodies, prefix matching, type-filtered
- Wiki links autocompleting against existing objects, creating lazily on first write
- Backlinks with context, and unlinked mentions offered

PORTABILITY, WHICH MATTERS MORE HERE THAN USUAL
- Markdown files on disk with the fields in frontmatter. **The type system must be expressible in plain files** or the whole archive depends on your software existing
- The index rebuildable from the files, always
- Export is just the folder

WHAT MATTERS MOST
Typing a line without leaving the writing, and automatic gathering on each object. Together they mean organisation happens as a by-product of writing rather than as a separate chore.