• Supports the latest TOML release (v1.0.0), plus optional support for some unreleased TOML features
  • Supports serializing to JSON and YAML
  • Proper UTF-8 handling (incl. BOM)
  • C++17 (plus some C++20 features where available, e.g. experimental support for char8_t strings)
  • Header-only (optional!)
  • Doesn't require RTTI
  • Works with or without exceptions
  • Tested on Clang (6+), GCC (7+) and MSVC (VS2019)
  • Tested on x64, x86 and ARM

API documentation

You're looking at it! Browse the docs using the links at the top of the page. You can search from anywhere by pressing the TAB key.

Basic examples

Parsing files

Call toml::parse_file() and work with the toml::table you get back, or handle any toml::parse_error that gets thrown:

#include <iostream>
#include <toml++/toml.h>

int main(int argc, char** argv)
   toml::table tbl;
       tbl = toml::parse_file(argv[1]);
           std::cout << tbl << "\n";
   catch (const toml::parse_error& err)
       std::cerr << "Parsing failed:\n" << err << "\n";
       return 1;

   return 0;

Parsing strings and iostreams

Call toml::parse() and work with the toml::table you get back, or handle any toml::parse_error that gets thrown:

Try this code on Compiler Explorer

#include <iostream> #include <sstream> #include <toml++/toml.h> using namespace std::string_view_literals; int main() { static constexpr std::string_view some_toml = R"( [library] name = "toml++" authors = ["Mark Gillard <>"] cpp = 17 )"sv; try { // parse directly from a string view: { toml::table tbl = toml::parse(some_toml); std::cout << tbl << "\n"; } // parse from a string stream: { std::stringstream ss{ std::string{ some_toml } }; toml::table tbl = toml::parse(ss); std::cout << tbl << "\n"; } } catch (const toml::parse_error& err) { std::cerr << "Parsing failed:\n" << err << "\n"; return 1; } return 0; }
authors = [ 'Mark Gillard <>' ]
cpp = 17
name = 'toml++'

authors = [ 'Mark Gillard <>' ]
cpp = 17
name = 'toml++'

Handling errors without exceptions

Can't (or won't) use exceptions? That's fine too. You can disable exceptions in your compiler flags and/or explicitly disable the library's use of them by setting the option TOML_EXCEPTIONS to 0. In either case, the parsing functions return a toml::parse_result instead of a toml::table:

#include <iostream>

#define TOML_EXCEPTIONS 0 // only necessary if you've left them enabled in your compiler
#include <toml++/toml.h>

int main()
   toml::parse_result result = toml::parse_file("configuration.toml");
   if (!result)
       std::cerr << "Parsing failed:\n" << result.error() << "\n";
       return 1;

   do_stuff_with_your_config(std::move(result).table()); // 'steal' the table from the result
   return 0;

Custom error formatting

The examples above use an overloaded operator<< with ostreams to print basic error messages, and look like this:

Error while parsing key: expected bare key starting character or string delimiter, saw '?'
    (error occurred at line 2, column 5)

In order to keep the library as small as possible I haven't bent over backwards to support things like custom colouring of the text in TTY environments, et cetera. That being said, the library provides the requisite information for you to build these yourself if necessary via toml::parse_error's source() and description() members:

toml::table tbl;
   tbl = toml::parse_file("configuration.toml");
catch (const toml::parse_error& err)
       << "Error parsing file '" << *err.source().path
       << "':\n" << err.description()
       << "\n  (" << err.source().begin << ")\n";
   return 1;

Working with TOML data

A TOML document is a tree of values, arrays and tables, represented as the toml::value, toml::array and toml::table, respectively. All three inherit from toml::node, and can be easily accessed via the toml::node_view:

Try this code on Compiler Explorer

#include <iostream> #include <toml++/toml.h> using namespace std::string_view_literals; int main() { static constexpr auto source = R"( str = "hello world" numbers = [ 1, 2, 3, "four", 5.0 ] vegetables = [ "tomato", "onion", "mushroom", "lettuce" ] minerals = [ "quartz", "iron", "copper", "diamond" ] [animals] cats = [ "tiger", "lion", "puma" ] birds = [ "macaw", "pigeon", "canary" ] fish = [ "salmon", "trout", "carp" ] )"sv; toml::table tbl = toml::parse(source); // different ways of directly querying data std::optional<std::string_view> str1 = tbl["str"].value<std::string_view>(); std::optional<std::string> str2 = tbl["str"].value<std::string>(); std::string_view str3 = tbl["str"].value_or(""sv); std::string& str4 = tbl["str"].ref<std::string>(); // ~~dangerous~~ std::cout << *str1 << "\n"; std::cout << *str2 << "\n"; std::cout << str3 << "\n"; std::cout << str4 << "\n"; // get a toml::node_view of the element 'numbers' using operator[] auto numbers = tbl["numbers"]; std::cout << "table has 'numbers': " << !!numbers << "\n"; std::cout << "numbers is an: " << numbers.type() << "\n"; std::cout << "numbers: " << numbers << "\n"; // get the underlying array object to do some more advanced stuff if (toml::array* arr = numbers.as_array()) { for (toml::node& elem : *arr) { // visitation helps deal with the polymorphic nature of TOML data elem.visit([](auto&& el) noexcept { if constexpr (toml::is_number<decltype(el)>) (*el)++; else if constexpr (toml::is_string<decltype(el)>) el = "five"sv; }); } // arrays are very similar to std::vector arr->push_back(7); arr->emplace_back<toml::array>(8, 9); std::cout << "numbers: " << numbers << "\n"; } // node-views can be chained to quickly query deeper std::cout << "cats: " << tbl["animals"]["cats"] << "\n"; std::cout << "fish[1]: " << tbl["animals"]["fish"][1] << "\n"; // can also be retrieved via absolute path std::cout << "cats: " << tbl.at_path("animals.cats") << "\n"; std::cout << "fish[1]: " << tbl.at_path("[1]") << "\n"; // ...even if the element doesn't exist std::cout << "dinosaurs: " << tbl["animals"]["dinosaurs"] << "\n"; //no dinosaurs :( return 0; }
hello world
hello world
hello world
hello world
table has 'numbers': 1
numbers is an: array
numbers: [ 1, 2, 3, 'four', 5.0 ]
numbers: [ 2, 3, 4, 'five', 6.0, 7, [ 8, 9 ] ]
cats: [ 'tiger', 'lion', 'puma' ]
fish[1]: 'trout'
cats: [ 'tiger', 'lion', 'puma' ]
fish[1]: 'trout'

Serializing as TOML, JSON and YAML

All toml++ data types have overloaded operator<< for ostreams, so 'serializing' a set of TOML data to actual TOML is done just by printing it to an ostream. Converting it to JSON and YAML is done in much the same way, but via a toml::json_formatter and toml::yaml_formatter.

Try this code on Compiler Explorer

#include <iostream> #include <toml++/toml.h> int main() { auto tbl = toml::table{ { "lib", "toml++" }, { "cpp", toml::array{ 17, 20, "and beyond" } }, { "toml", toml::array{ "1.0.0", "and beyond" } }, { "repo", "" }, { "author", toml::table{ { "name", "Mark Gillard" }, { "github", "" }, { "twitter", "" } } }, }; // serializing as TOML std::cout << "###### TOML ######" << "\n\n"; std::cout << tbl << "\n\n"; // serializing as JSON using toml::json_formatter: std::cout << "###### JSON ######" << "\n\n"; std::cout << toml::json_formatter{ tbl } << "\n\n"; // serializing as YAML using toml::yaml_formatter: std::cout << "###### YAML ######" << "\n\n"; std::cout << toml::yaml_formatter{ tbl } << "\n\n"; return 0; }
###### TOML ######

cpp = [ 17, 20, 'and beyond' ]
lib = 'toml++'
repo = ''
toml = [ '1.0.0', 'and beyond' ]

github = ''
name = 'Mark Gillard'
twitter = ''

###### JSON ######

    "author" : {
        "github" : "",
        "name" : "Mark Gillard",
        "twitter" : ""
    "cpp" : [
        "and beyond"
    "lib" : "toml++",
    "repo" : "",
    "toml" : [
        "and beyond"

###### YAML ######

  github: ''
  name: 'Mark Gillard'
  twitter: ''
  - 17
  - 20
  - 'and beyond'
lib: toml++
repo: ''
  - '1.0.0'
  - 'and beyond'

Speeding up compilation

Because toml++ is a header-only library of nontrivial size you might find that compilation times noticeably increase after you add it to your project, especially if you add the library's header somewhere that's visible from a large number of translation units. You can counter this by disabling header-only mode and explicitly controlling where the library's implementation is compiled.

Step 1: Set TOML_HEADER_ONLY to 0 before including toml++

This must be the same everywhere, so either set it as a global #define in your build system, or do it manually before including toml++ in some global header that's used everywhere in your project:

// global_header_that_includes_toml++.h

#include <toml.hpp>

Step 2: Define TOML_IMPLEMENTATION before including toml++ in one specific translation unit

// some_code_file.cpp

#include "global_header_that_includes_toml++.h"

Bonus Step: Disable any library features you don't need

Some library features can be disabled wholesale so you can avoid paying their the compilation cost if you don't need them. For example, if all you need to do is serialize some code-generated TOML and don't actually need the parser at all you, can set TOML_ENABLE_PARSER to 0 to disable the parser altogether. This can yield fairly significant compilation speedups since the parser accounts for a good chunk of the library's code.

Adding toml++ to your project

The library comes in two flavours, 🍦️ Single-header and 🍨️ Regular. The API is the same for both.

"The old fashioned way"

Clone the repository from GitHub, and then:

🍦️ Single-header flavour

  1. Drop toml.hpp wherever you like in your source tree
  2. There is no step two

🍨️ Regular flavour

  1. Add tomlplusplus/include to your include paths
  2. #include <toml++/toml.h>


Add tomlplusplus/3.0.1 to your conanfile.


Add tomlpp to your package.json5, e.g.:

depends: [


meson wrap install tomlplusplus

You can install the wrap with: After that, you can use it like a regular dependency: ‘tomlplusplus_dep = dependency('tomlplusplus’)`. You can also add it as a subproject directly.


vcpkg install tomlplusplus

CMake FetchContent

    GIT_TAG        v3.0.1

Other environments and package managers

toml++ is a fairly new project and I'm not up-to-speed with all of the available packaging and integration options in the C++ ecosystem. I'm also a cmake novice, for better or worse. If there's an integration option missing be assured that I fully support it being added, and welcome pull requests!

Special mention: Python

There exists a python wrapper library built around toml++ called pytomlpp which is, at the time of writing, the only natively-compiled TOML library available for python, and thus much faster than many of the alternatives:

Parsing data.toml 5000 times:
  pytomlpp:   0.694 s
     rtoml:   0.871 s ( 1.25x)
     tomli:   2.625 s ( 3.78x)
      toml:   5.642 s ( 8.12x)
     qtoml:   7.760 s (11.17x)
   tomlkit:  32.708 s (47.09x)

Install it using pip:

pip install pytomlpp

Note that I'm not the owner of that project so if you wish to report a bug relating to the python implementation, please do so at their repository, not on the main toml++ one.

Library configuration options

The library exposes a number of configuration options in the form of compiler #defines. Things like changing the optional<T> type, disabling header-only mode, et cetera. The full list of configurables can be found on the Library Configuration page.


Contributions are very welcome! Either by reporting issues or submitting pull requests. If you wish to submit a pull request, please see CONTRIBUTING for all the details you need to get going.


toml++ is licensed under the terms of the MIT license - see LICENSE.

If you're using the single-header version of the library you don't need to explicitly distribute the license file; it is embedded in the preamble at the top of the header.

Contacting the author

For bug reports and feature requests please use the Github Issues system. For anything else you're welcome to reach out via other means. In order of likely response speed: