TOML ↔ JSON Converter

About the TOML ↔ JSON Converter

Turn a TOML file — a Cargo.toml, a pyproject.toml, a Hugo or Netlify config — into JSON or YAML, or write TOML from JSON or YAML. Everything runs in your browser and nothing is rounded: integers keep every digit (9223372036854775807 stays exactly that), floats keep the digits you wrote, and date-times keep their offset and every digit of the fraction (1979-05-27T00:32:00.999999-07:00).

When a TOML file is invalid, you get the rule that was broken and where: "Table [server] is defined twice: it was first defined on line 3", "The dotted key b.c.t cannot add to b.c: it is a table (its [header] is on line 1)…", "Invalid time: the hour is 24, but hours go from 00 to 23". The parser follows TOML 1.1.0 and, on request, the stricter TOML 1.0.0, and it passes every applicable test of the official toml-test suite.

Common uses

• Reading a Cargo.toml, pyproject.toml or netlify.toml with jq or a script that only understands JSON.
• Moving a configuration between tools that use TOML, JSON or YAML (Hugo accepts all three).
• Finding out exactly why a TOML file is rejected — duplicate keys, a table defined twice, a dotted key that collides with a header.
• Checking that a file avoids TOML 1.1-only syntax before a TOML 1.0 parser reads it.

How TOML values become JSON and YAML

• Strings stay strings; multi-line strings keep their line breaks (a CRLF inside one becomes \n).
• Integers keep every digit at any size; 0x, 0o and 0b integers become decimal. JSON readers built on JavaScript's JSON.parse round integers beyond ±9,007,199,254,740,991, so the converter tells you when there are any and can write them as strings instead.
• Floats keep the digits as written (+1.0 becomes 1.0, 1_000.5 becomes 1000.5). JSON has no inf or nan, so they become null with a warning; YAML writes .inf and .nan.
• Offset date-times, local date-times, local dates and local times become strings in RFC 3339 form: the offset and the fraction are kept, the separator becomes T, z becomes Z, and a time written without seconds (allowed in TOML 1.1) gets :00. In YAML they are written unquoted, so YAML 1.1 parsers such as PyYAML read them as timestamps; local times are quoted because YAML 1.1 reads 17:32:00 as a number.
• Tables, inline tables and arrays of tables become JSON objects and arrays, in the order the keys were written.

Typed JSON writes every value as {"type": …, "value": …}, the format of the toml-test suite, so nothing about the TOML type is lost.

TOML 1.1 and TOML 1.0

TOML 1.1.0 added five things to 1.0.0: line breaks, comments and a trailing comma inside inline tables, the \e escape, the \xHH escape, and times without seconds (07:32). Everything else is the same, so every TOML 1.0 file is also a TOML 1.1 file.

In TOML 1.1.0 mode each use of the new syntax is listed as a note, because parsers that only read TOML 1.0 reject the file — Python's tomllib reads TOML 1.0 up to Python 3.14, for example. In TOML 1.0.0 mode they are errors that tell you the 1.0 way to write them. TOML written by this converter (from JSON or YAML) never uses 1.1-only syntax, so TOML 1.0 parsers can read it too; an integer beyond 64 bits, which many parsers reject, is flagged with a warning.

Rules the converter checks

• A key can be defined only once in a table, and "key", 'key' and key are the same key.
• A table can be defined only once. [a.b] implies [a], which can still be defined once later.
• Dotted keys (fruit.color = "red") create tables that cannot get a [header] later, and dotted keys cannot add to a table that has a header — but [fruit.apple.texture]-style sub-tables are allowed.
• Inline tables ({ … }) are complete: nothing can be added to them later.
• [[name]] cannot add to an array written as a value (name = []), and a key cannot be both a table and an array of tables.
• Bare keys use only A–Z a–z 0–9 _ -; strings must be quoted; numbers have no leading zeros; underscores sit between digits; dates must exist (no 30 February) and times must be valid.

These rules come from the TOML specification and are checked against all of the official toml-test cases.

Writing TOML from JSON or YAML

• TOML has no null: a null value is an error that names its path, or, with Leave out null values, the key is skipped (nulls inside arrays are always errors, since arrays cannot skip items).
• A TOML document is a table, so the JSON or YAML must be an object; a top-level array has to be put under a key first. A YAML stream with several --- documents must be converted one document at a time.
• With Dates and times as TOML dates, strings that are exact RFC 3339 dates or times (2024-03-14, 09:30:00, 2024-03-14T09:30:00+05:30) become TOML date values. In YAML only unquoted values are converted; quoted ones stay strings.
• In each table, plain values come before sub-tables (TOML requires this), so the order of keys can differ from the input; the data is the same.

Limitations

• Comments are not carried over: JSON has none, and comments in YAML input are dropped.
• This is a converter and checker, not a TOML formatter: TOML output is written fresh, without the original comments or layout.
• Integers outside the 64-bit range (−9,223,372,036,854,775,808 to 9,223,372,036,854,775,807) are kept exactly with a warning, but many TOML parsers reject them.
• It checks TOML syntax, not the meaning of a particular file: it does not know which keys Cargo, pip or Hugo expect.
• Tables and arrays nested more than 500 levels deep — counting [headers], dotted keys and inline tables — are refused, in both directions.

Privacy

Everything happens in your browser. What you enter or open here is not uploaded or stored by MySmartCoPilot.