General Comments

This paper presents a suit of tools related to the forward and inverse modelling of mechanical anisotropy. The capabilities which these tools provide are in my opinion a very important contribution to scientific progress within the scope of this journal. The paper described the purpose of the individual tools and explains at a high level what kind of changes level changes which where made (in case of being build on previous work), some considerations on how to use it and shows example explanations. 

Unfortunately, there are, in my opinion, two major sections missing in this paper. The first sections is a proper methods section, which describes for each tool how it works. For example, what are the equations which are being solved. If it is based on previous work (such as D-REX), what changes have been made to the methods (more detailed then the current description) and how does this align with the assumptions made in the original code. What values of different parameters are used for computing for example wadsleyite or bridgmanite (both in D-REX and in the computation of the elastic tensor).

A second section which is missing is a benchmark section, to show that what is stated in the methods section actually works as intended. These should be at least a small suite of simple tests. For example computing the CPO in a simple shear environment, which can be matched against analytic results, the results of other codes and/or experiments. The only benchmarks which are shown are performance benchmarks.

Because these two sections are missing it is not possible to properly review the paper, since the authors do not show what they have done exactly and that they have done it correctly. Although the code is open-source, and it could in theory be checked by looking into the code and creating and running benchmarks yourself, I do think this should be part of the paper.

Specific comments

- The geodynamics code which are used for the examples should  be mentioned. Looking at the references, it seems like for some of them I3MG is used, but this is not explicitly mentioned and it is not clear for all of them.

- Github is not a software archive, since it can easily be removed or changed. It is good to mention it, but you will also need a doi of for the software. You can get this for example from Zenodo.

- The repository has a license (MIT), a changlog document, a proper versioning scheme and a comprehensive user manual. I assume that this paper is about version 2.0, but this should be explicitly stated in the paper. I could not find any tests or benchmarks in the repository, although it could be that the cookbooks function as tests.

Technical corrections

line 135-137: Name and cite studies which are used together with the lookup tables. Also, it just states that the fabric selection is P-T dependent, so no water dependence? To come back to the general point, it needs to be explained how these fabrics are selected. A graph or table would be nice, if feasible.

line 361-362: Workstation is a vague term, mention CPU and ram needed to do that that calculation in the stated time.

line 428: I assume that the authors mean spreading memory load over several nodes, not CPU's.