• Latest version: qgraf-3.6.5 (October 2022).
  Please report any error you may happen to find.

• Programming language: Fortran 2008 for versions released since 2020 (qgraf-3.5+), and Fortran 77 for earlier versions.
  Executables/binaries are not distributed, the program has to be compiled and linked.
  Employing GNU Fortran for that task should be straightforward, eg there should be no need to specify a Fortran standard (see eg file qgraf-3.6.5.pdf). There are executable/binary versions of GFortran for several operating systems, as described in the GFortranBinaries webpage.

• The features added in the last three versions include (eg):
  selecting 1–vertex irreducible diagrams (3.4);
  selecting diagrams with non-factorizable cycle spaces (3.4);
  generating multiple output files in the same run (3.5);
  inputing simple command-line arguments (3.5);
  an extended language for describing input models (3.6);
  a more general loops statement (3.6).

• Current plans: qgraf-3.4 should be available and supported for some time yet (minimal fixes as need be); qgraf-3.6 is now a stable version too; the Fortran and C interfaces should appear in the next version.

Qgraf is based on the method described in the following paper:

• Automatic Feynman graph generation
  J. Comput. Phys. 105 (1993) 279–289
  https://doi.org/10.1006/jcph.1993.1074
  

  ▻ show outline ▻ hide
  This paper presents a concise description of an orderly algorithm for generating Feynman diagrams, on which Qgraf is based. In addition, it presents a diverse set of numbers — either the numbers of graphs in certain sets or the sums of the respective symmetry factors — which may be employed to cross-check the correctness of Feynman diagram generators.

The diagram generation with explicitly mixed propagators (kinetic mixing) can be reduced to the usual one:

• Feynman graph generation and propagator mixing, I
  Comput. Phys. Commun. 269 (2021) 108103
  https://doi.org/10.1016/j.cpc.2021.108103
  

  ▻ show outline ▻ hide
  Given a QFT model A with explicitly mixed propagators, this paper shows how to construct a model A^⁎ without that kind of mixing, such that the Feynman diagrams in A may be easily derived from the diagrams in A^⁎. Moreover, it explains in detail why the method works; for example, it shows that if a diagram d^⁎ in A^⁎ gives rise to some diagram d in A then d and d^⁎ have identical symmetry factors.