CatGo (open-source AI-agent workbench): native CP2K input generation, analysis & post-processing

[Guangsheng Liu]

  Dear CP2K community,

  Our group built CatGo, an open-source AI-agent-driven workbench for
  computational chemistry focused on catalysis. It combines interactive
  3D structure building/editing, a visual DAG workflow editor, HPC job
  submission/monitoring, and a natural-language assistant in one tool.

    CatGo:    https://github.com/Hello-QM/catgo-LRG
    Web app:  https://app.catgo-ucsd.org   (no install)
    Docs:     https://docs.catgo-ucsd.org
    Preprint: https://doi.org/10.26434/chemrxiv.15002984/v1

  CP2K is a first-class, natively supported engine in CatGo. From a
  structure built or edited in the 3D viewer you can:

    - generate CP2K inputs (GPW method, GTH pseudopotentials);
    - chain single_point / geo_opt / cell_opt / MD (and more) as nodes
      in a visual workflow;
    - submit to HPC (SLURM/PBS/LSF/SGE) and monitor jobs live;
    - parse and post-process CP2K results in the same UI (energies,
      relaxation/MD trajectories, DOS/PDOS).

  We are reaching out for two reasons:

    1. A sincere thank-you - CP2K is a core engine for our catalysis
       users, especially for large slab/surface systems.
    2. To open the door to collaboration. We would value feedback from
       the CP2K developers on our input generation and defaults, and
       pull requests / joint work are very welcome - for example
       improving the structure -> CP2K input path for surfaces and
       large catalysis systems, and the parsing/post-processing layer.

  Thank you for CP2K.

  Best regards,
  The CatGo team