Get Started

MNE-CPP can be used in three ways. Pick the one that matches what you want to do:

You want to…	Start here
Run the applications — record, browse, analyze and visualize data	Download an installer, or open MNE Browse or MNE Inspect in the browser
Write C++ against the libraries	This page: build, install and link MNE-CPP
Understand the methods	The processing workflow, forward modeling and inverse methods

The rest of this page builds MNE-CPP itself, installs it, and then builds a separate program that uses it as a library — the two things are distinct: you only build MNE-CPP once, and your own projects find it with CMake's find_package.

1. Build MNE-CPP

You need Git, CMake and a C++20 compiler (GCC 13+, Apple Clang or MSVC); the build guide lists the tested versions and alternatives. The init script downloads the Qt and Eigen builds that CI uses into src/external/ and configures build/developer-dynamic:

Linux and macOS
git clone --recursive https://github.com/mne-tools/mne-cpp.git
cd mne-cpp
./init.sh
cmake --build build/developer-dynamic --parallel 4

On Windows run .\init.bat and cmake --build build\developer-dynamic --config Release --parallel 4 instead. Check the build by running an example:

./out/Release/examples/ex_math

--recursive also fetches mne-cpp-test-data into resources/data/mne-cpp-test-data, the small recordings that the tests and many examples read. In an existing clone, get it with git submodule update --init resources/data/mne-cpp-test-data.

2. Install the libraries

Install headers, libraries and the CMake package files into a prefix of your choice:

cmake --install build/developer-dynamic --prefix "$HOME/mne-cpp-sdk"

The prefix now contains include/<library>/…, the shared libraries, and lib/cmake/MNE-CPP/MNE-CPPConfig.cmake, which defines one imported target per library — MNE-CPP::mne_fiff, MNE-CPP::mne_dsp, MNE-CPP::mne_inv and so on.

3. Write a program that uses MNE-CPP

Create a new directory outside the MNE-CPP checkout with two files. The program opens a FIFF recording with FiffRawData, prints its size and reads the first second of every channel with read_raw_segment.

raw_summary/CMakeLists.txt
cmake_minimum_required(VERSION 3.21)
project(raw_summary LANGUAGES CXX)

set(CMAKE_CXX_STANDARD 20)
set(CMAKE_CXX_STANDARD_REQUIRED ON)

find_package(MNE-CPP 2.4 REQUIRED)

add_executable(raw_summary main.cpp)
target_link_libraries(raw_summary PRIVATE MNE-CPP::mne_fiff)

raw_summary/main.cpp
#include <fiff/fiff_raw_data.h>

#include <QFile>

#include <cstdio>

int main(int argc, char* argv[])
{
    if (argc < 2) {
        std::fprintf(stderr, "usage: %s <raw.fif>\n", argv[0]);
        return 1;
    }

    QFile file(argv[1]);
    FIFFLIB::FiffRawData raw(file);
    if (raw.isEmpty()) {
        std::fprintf(stderr, "could not read %s\n", argv[1]);
        return 1;
    }

    const double sfreq = raw.info.sfreq;
    std::printf("%d channels at %.1f Hz, %.1f s\n", raw.info.nchan, sfreq,
                (raw.last_samp - raw.first_samp + 1) / sfreq);

    // The first second of every channel: rows are channels, columns samples.
    Eigen::MatrixXd data, times;
    if (!raw.read_raw_segment(data, times, raw.first_samp, raw.first_samp + static_cast<int>(sfreq) - 1)) {
        return 1;
    }
    std::printf("read %lld x %lld samples, first channel %s\n", static_cast<long long>(data.rows()),
                static_cast<long long>(data.cols()), qPrintable(raw.info.ch_names.first()));
    return 0;
}

Configure it with the install prefix, Qt and Eigen on CMAKE_PREFIX_PATH. With the dependencies that init downloaded, that is (replace ~/mne-cpp with your checkout):

cmake -S raw_summary -B raw_summary/build -DCMAKE_BUILD_TYPE=Release \
      -DCMAKE_PREFIX_PATH="$HOME/mne-cpp-sdk;$HOME/mne-cpp/src/external/qt/dynamic;$HOME/mne-cpp/src/external/eigen"
cmake --build raw_summary/build
./raw_summary/build/raw_summary ~/mne-cpp/resources/data/mne-cpp-test-data/MEG/sample/sample_audvis_trunc_raw.fif

The last two lines of the output are:

376 channels at 300.3 Hz, 20.0 s
read 376 x 300 samples, first channel MEG0113

The lines before them are the reader's own log: the projectors stored in the file and the sample range. If you use a Qt from the Qt Online Installer or a system Eigen, put those prefixes on CMAKE_PREFIX_PATH instead.

Where this was verified

These steps were run end to end on macOS with the Qt and Eigen bundles that init.sh downloads. On Windows the same CMake project works with MSVC; the Qt bin directory must be on PATH when you run the program, as described in the build guide.

Next steps

Find an example by task, recording type or library in the examples explorer.
Look up a class in the API reference, or press Ctrl K to search for it by name.
Follow a complete analysis in the processing workflow, from raw data to source estimates.

1. Build MNE-CPP​

2. Install the libraries​

3. Write a program that uses MNE-CPP​

Next steps​

1. Build MNE-CPP

2. Install the libraries

3. Write a program that uses MNE-CPP

Next steps