This manuscript introduces a newly developed modelling framework copan:LPJmL, which can simulate a wide range of socio-ecological dynamics. This work makes an important contribution to addressing the conceptual and technical challenges in modelling intertwined human and nature processes. The overall structure of the manuscript is clearly presented. The three examples introduced well highlight this framework’s flexibility and versatility. However, the manuscript could be improved by distributing the content more properly between the manuscript and the technical manual to enhance the ease of understanding for readers. In general, I would suggest leaving the conceptual content, modelling decisions/justifications, and model logic in the manuscript, while the technical specifications in the GitHub/Zenodo repo, so that readers can easily navigate from high-level conceptual description to implementation details. More suggestions follow here.

1. copan:LPJmL is defined as a framework in the manuscript. It might be ambiguous to readers as a framework can mean different things, such as a conceptual framework, a model, a sub-model, or a wrapper for the existing LPJmL to interface with copan:CORE. It would be best to clarify what copan:LPJmL exactly is early in the text. This could also help readers know clearly what kind of contribution the manuscript is intended to stress.
2. Usually, readers are motivated to invest efforts in looking into technical details (e.g., code) only if the concepts are attractive and clearly presented. A class diagram appears unable to serve as an efficient means of communication, as it normally provides neither conceptual simplicity for broad readers nor accurate technical details for developers. It would be best to replace the class diagram with a conceptual framework that illustrates the connections and information flow between model components.
3. Accordingly, much of the technical description, especially the parts that mix narratives with variable names defined in the code, could be organised into a README.md/txt file in the code repo, leaving only the description of the conceptual idea and model logic (using text, pseudocode, or equations) in the paper. I saw the current code repo only has a minimal technical description in contrast with that in the manuscript. A proper redistribution of the content might be necessary.
4. The code shown in the figures seems to provide limited information about the model mechanism. If the code only serves as examples of the use of copan:LPJmL, the README file in the repo should be the best place.
5. The manuscript could be improved by avoiding relying on Python syntax in the narrative. Although Python and object-oriented programming are widely used in research, using syntax-dependent expressions like “world.input[1]” might not be an ideal way to express the gist of copan:LPJmL.
6. It is fantastic to learn that copan:LPJmL can accommodate a range of modelling approaches. However, the three examples do not contain enough information about how this framework can achieve this. Developers or model users might not be clear about what efforts they should make to switch modelling approaches within copan:LPJmL. I would suggest describing the examples using an identical structure: for each example, including background information, concepts, model processes, model settings, and outcomes, while highlighting the role of copan:LPJmL and what model users should do. In addition, please cite the source code of each example in the text.
7. Line 287, the citation “Schwarz et al.” is not complete.
8. Line 370, “)” is missed in “(Chapter 3.1”.  In addition, I am not sure whether “section” or “chapter” is more accurate in the context. Please check the convention regarding the word choice of GMD.
9. Please cite the sources of the figures and code if they are already publicly available elsewhere.

This article describes the coupling of the copan:CORE World-Earth model and the DGVM LPJmL, giving three examples of its use for agent-based, rule-based and LLM-based modelling. It’s great to see this approach being developed further and the resulting copan:LPJmL framework has real promise for modelling the interactions of human and natural (land) systems, and doing so in ways that are distinct from established models. The easy availability of the framework is also important, and it should be useful to many.

While I think the article is strong, there are several points that either could or need to be strengthened in a revision. These include conceptual descriptions of some aspects of the approach, but also how some of the core functionalities are operationalized. I’ve highlighted these in the comments below.  

General comments

• The novel capabilities of the model are given in overview, but the conceptual details aren’t always apparent. There’s an emphasis on technical description (largely appropriate given the journal) but this isn’t always convincingly linked back to the objectives. In particular more content on how (and which) social and decision processes can be modelled would be really useful. There are inevitably some parts that are unclear at the moment, but also I think some under-selling in that the authors leave more open questions about applicability than they need to. See specific comments below on these points.
• Resolution: The default resolution given here is fairly coarse for some of the processes (especially social) that the authors highlight as being important; things like social norms and networks, but also diverse vegetation growth within what are represented as single land holdings. Some more detail and justification of the implications of resolution, and how they can be accounted for in an applied model, is needed.
• Figures: I’m not convinced how useful the figures comprising code (or components, as in Fig. 2) are. The code is already available, so it might be best to use more generic/descriptive representations here to give conceptual overviews and to provide more interpretation of model capabilities.
• The discussion would also benefit from substantially more reflection on applicability and interpretation. Some of this could perhaps deal with flexibility in the model to represent more things or to represent them differently. Even the authors seem unconvinced at times about this!

Specific comments   

Lines 21-22: ‘Committee, 1986’ – incomplete reference?

Lines 28-29: It might be worth stating how you see WEM relating to social-ecological systems modelling

Lines 83-84: Repeated ‘detailed’

Line 88: the reference ‘Schwarz et al.’ doesn’t have a year

Line 109: It’s a bit unclear what ‘the framework’ is.

Line 110: Clarify the ‘aforementioned taxonomy’ by writing out the terms again?

Line 113: Is the word ‘entities’ being used in the same sense throughout the description? It seems imprecise in its meaning.

Lines 115-116: Is the Group entity distinct from the social institutions mentioned previously?

Lines 126-127: It would be useful to have here a description or link to one, of what the various agricultural management practices can be.

Lines 129-130: Some text already at this point on how this resolution aligns with the objectives of the model would be helpful.

Figure 2: I’m particularly unsure how helpful this Figure is. There are lots of terms in there that aren't (and can't easily be) interpreted, so add little value. This might be better in SI with full explanations, and a simpler version here with normal language explanations.

Line 164: ‘Serves as a’ or ‘provides a’ alternatively.

Line 170: ‘are only sent…’

Lines 195-199: It would be useful to have less technical (more conceptual) summaries above this point to establish that these things can be modelled meaningfully, beyond the technical capacity of the model. But even at that technical level, it’s not clear to me how decisions and interactions can/should be modelled at this resolution.

Lines 205-206: Are land use and agricultural management given as examples of social processes, or ENV interactions & feedbacks? The following list about the CUL taxon does I think need to be justified and explained better.

Table 1: The ‘Key variables’ for Land Use seem questionable – what is included in ‘land systems’?

Table 2: Should the vegetation row include PFTs?

Lines 256-257: I think it would be very worthwhile to describe how this can be extended.

Figure 5: legend – why ‘following’ Fig 8? (or, why in this order). Also are crop yields and soil carbon the only things observed by the farmer types, and are they known perfectly?

Lines 266-267: As in previous comment, it seems unnecessarily unrealistic for farmer agents to know these things perfectly in the modelled world, and not to know about other constraints or outcomes. Explain/extend?

Line 273: Can the decision-making process be varied?

Lines 288-289: Anything you can say about how they can do this would be really interesting here.

Lines 291-292: This is a very underwhelming end to the example! Surely you can – and need to – argue that they can at least generate some sort of new insight?

Figure 8 legend: This description seems quite distinct from real-world decisions, and more of a forward-looking optimisation. I’m not sure how much or easily this can be varied to represent alternative ways of making decisions, but it seems important that it can be.

Figure 9: It’s hard to see the value of this figure.

Line 381-382: It’s good that the reasoning is provided as this can change the decision. Is this checked by a human user?

Line 424 on: I would prefer more reflection than a description followed by 'shortcomings'. The model of course can't do everything, so the framing here seems overly-negative, but then interpretation of what the model actually can do is largely missing.

Summary and Outlook: This is quite repetitive of the first part of the discussion; I think you could save space here and cover more ground overall, especially guiding the reader in what the technical capabilities of the model can be used for.