Manual installation of agama in a plain conda enviroment

install_agama.sh

# create env and install dependency
conda create -n agama python=3.8 numpy gsl glpk eigen cvxopt
conda activate agama

# download files
cd $HOME/src
wget https://github.com/GalacticDynamics-Oxford/Agama/archive/master.zip -O agama.zip
wget https://github.com/GalacticDynamics-Oxford/unsio/archive/master.zip -O unsio.zip
wget https://github.com/cvxopt/cvxopt/archive/master.zip -O cvxopt.zip
unzip agama.zip
unzip unsio.zip
unzip cvxopt.zip
mv Agama-master agama
mv unsio-master unsio
mv cvxopt-master cvxopt

# compile unsio
cd unsio
make
cd ..

# check the enviroment path
echo $CONDA_PREFIX
python -c "import numpy as np; print(np.get_include())"
python -c "import sysconfig as sc; print(sc.get_paths()['include'])"

# create makefile
cat <<EOF >agama/Makefile.local
# ---------------------------------------------------
# Makefile.local
FC = gfortran
LINK = \$(CXX)

# optional: -std=c++11
CXXFLAGS += -fPIC -fopenmp -Wall -O2 -march=native

COMPILE_FLAGS += -I\$(CONDA_PREFIX)/include/
LINK_FLAGS    += -L\$(CONDA_PREFIX)/lib -lgsl -lgslcblas

COMPILE_FLAGS += -DHAVE_PYTHON -I\$(CONDA_PREFIX)/include/python3.8 -I\$(CONDA_PREFIX)/lib/python3.8/site-packages/numpy/core/include
LINK_FLAGS    += -L\$(CONDA_PREFIX)/lib -lpython3.8

COMPILE_FLAGS += -DHAVE_EIGEN -I\$(CONDA_PREFIX)/include/eigen3
COMPILE_FLAGS += -DHAVE_CVXOPT -I\$(HOME)/src/cvxopt/src/C

COMPILE_FLAGS += -DHAVE_UNSIO -I\$(HOME)/src/unsio/unsio
LINK_FLAGS    += -L\$(HOME)/src/unsio -lunsio -lnemo

LINK_FLAGS    += -Wl,-rpath,\$(CONDA_PREFIX)/lib
EXE_FLAGS     += -Wl,-rpath,\$(CONDA_PREFIX)/lib  # necessary for compling exe
# ---------------------------------------------------
EOF

# make and test
cd agama
make  # note: do not use parallel make
make test

Then one may install other useful python packages afterwards

conda install ipython jupyter jupyterlab numpy scipy numba pandas astropy scikit-learn sympy cython matplotlib seaborn h5py