| name | makrell-data |
| description | Work with Makrell data and markup formats — MRON (.mron), MRML (.mrml), and MRTD (.mrtd). Use when creating, editing, validating, or explaining these formats. |
| allowed-tools | Read Grep Glob Edit Write Bash |
Makrell Data Formats
You are helping in a project that uses Makrell data and markup formats.
These formats share the MBF structural core with the Makrell language family.
Format Overview
| Format | Purpose | File ext | Closest analogue |
|---|
| MRON | Data notation (config, API) | .mron | JSON / TOML |
| MRML | Markup notation | .mrml | HTML / XML |
| MRTD | Tabular data | .mrtd | CSV / TSV |
MRON — Makrell Object Notation
MRON is a Makrell-shaped alternative to JSON. Lighter punctuation, same structural expressiveness.
Scalars
name "Rena Holm"
count 42
pi 3.14
active true
missing null
Strings require double quotes. Numbers, booleans (true/false), and null are bare.
Objects (top-level)
project "Makrell"
version "0.10.0"
active true
A top-level MRON document is implicitly an object — each line is a key/value pair.
Lists
tags ["docs" "formats" "languages"]
scores [95 87 91]
Square brackets, space-separated values.
Nested objects
site {
title "makrell.dev"
sections [
{
name "concepts"
public true
}
{
name "tutorials"
public true
}
]
}
Curly braces for nested objects. Lists of objects use [{ ... } { ... }].
Complete example
owner "Rena Holm"
last_update "2026-04-03"
books [
{
title "That Time of the Year Again"
year 1963
author "Norton Max"
}
{
title "One for the Team"
year 2024
author "Felicia X"
}
]
MRON vs JSON mapping
| JSON | MRON |
|---|
{ "key": "value" } | key "value" |
{ "a": [1, 2, 3] } | a [1 2 3] |
{ "x": { "y": 1 } } | x { y 1 } |
"string" | "string" |
42, true, null | 42, true, null |
| commas, colons | whitespace separation |
MRML — Makrell Markup Language
MRML is a Makrell-shaped alternative to HTML/XML. Tree structure expressed through nested curly forms.
Basic elements
{h1 Hello World}
{p This is a paragraph.}
The first token after { is the tag name. Everything else is content.
Attributes
{p [class="lead" id="intro"]
A paragraph with attributes.}
Square brackets after the tag name hold attributes.
Nested structure
{html
{head
{title A Test}}
{body
{h1 This is a Test}
{p [style="color: red"] Just some {b bold} text here.}}}
Inline elements
{p Use the {b shared concepts} section first.}
{p Visit {a [href="/docs"] the docs} for more.}
Complete page example
{html
{head
{meta [charset="utf-8"]}
{title My Page}
{link [rel="stylesheet" href="style.css"]}}
{body
{header
{nav
{a [href="/"] Home}
{a [href="/about"] About}}}
{main
{section [class="hero"]
{h1 Welcome}
{p One structural family for code, data, and markup.}}
{section [class="features"]
{div [class="card"]
{h2 Functional}
{p Pipes, operators, and composition.}}
{div [class="card"]
{h2 Multi-host}
{p Python, TypeScript, and .NET.}}}}
{footer
{p Built with Makrell.}}}}
MRML vs HTML mapping
| HTML | MRML |
|---|
<h1>Title</h1> | {h1 Title} |
<p class="x">Text</p> | {p [class="x"] Text} |
<div><p>A</p><p>B</p></div> | {div {p A} {p B}} |
<a href="/">Home</a> | {a [href="/"] Home} |
| Close tags | Closing } (structural nesting) |
MRTD — Makrell Tabular Data
MRTD is a whitespace-separated tabular format. Simpler than CSV for many common shapes.
Untyped (header + data)
name city
Ada Oslo
Ben Bergen
The first row is the header. Subsequent rows are data. Cells are whitespace-separated.
Typed columns
name:string age:int active:bool
Ada 32 true
"Rena Holm" 29 false
Header cells can carry type annotations: int, float, bool, string.
Quoted values
"full name":string city:string
"Rena Holm" "Bergen sentrum"
Quote values that contain spaces or are not valid identifiers.
Multiline rows
name:string note:string score:float
( "Rena Holm"
"line wrapped"
13.5 )
Wrap rows in ( ) when they need to span multiple lines.
Profiles (extended scalars)
when bonus
"2026-04-03"dt 3k
The extended-scalars profile adds suffixed literals like date (dt) and numeric shortcuts (k, M). Profiles are opt-in.
MRTD vs CSV mapping
| CSV | MRTD (untyped) | MRTD (typed) |
|---|
name,age,active | name age active | name:string age:int active:bool |
Ada,32,true | Ada 32 true | Ada 32 true |
"Rena Holm",29,false | "Rena Holm" 29 false | "Rena Holm" 29 false |
| Comma-separated | Whitespace-separated | Whitespace-separated |
| No type info | No type info | Type annotations in header |
Validation commands
All three formats can be checked via CLI:
makrellsharp check-mron file.mron --json
makrellsharp check-mrml file.mrml --json
makrellsharp check-mrtd file.mrtd --json
Programmatic access
Python (MakrellPy)
from makrell.mrtd import read_records
rows = read_records("""
name:string age:int active:bool
Ada 32 true
Ben 41 false
""")
TypeScript (MakrellTS)
import { parseMrtd, readMrtdRecords } from "makrellts";
const rows = readMrtdRecords(`
name:string age:int active:bool
Ada 32 true
Ben 41 false
`);
C# (Makrell#)
var rows = MrtdTyped.ReadRecords<Person>("""
name:string age:int active:bool
Ada 32 true
Ben 41 false
""");
Guidelines
When writing MRON:
- No commas or colons between keys and values — use whitespace
- Strings always use double quotes
- Top-level is implicitly an object
When writing MRML:
- First token after
{ is the element name
- Attributes go in
[...] immediately after the tag name
- Close with
}, not a closing tag — nesting is structural
When writing MRTD:
- First row is always the header
- Whitespace separates cells, not commas
- Quote any value containing spaces
- Type annotations are optional:
name and name:string are both valid headers
Documentation
