Prepare a native model input.

Describe your entities, or import an input file you already have. GreenFold maps protein chains to reviewed human UniProt entries, stages the precomputed MSAs, and returns a ZIP with a rewritten input and run commands.

Input

Entities

Build one model input below. Each Protein entity can start from a UniProt ID or a sequence; multiple protein entities become chains in the same complex. For repeated targets or supported batches, use Local runner & automation with the same model input semantics.

Each entity becomes one or more chains. Copies > 1 create additional chain IDs sharing the same sequence, the same way AlphaFold Server's "Copies" field works. Not sure why "Ion" sometimes ends up as a ligand? See the schema notes.

Output settings model filename
Add entities, or import an existing input.
Options for reverse sequence matching
Chains below identity 0.9 do not use precomputed GreenFold MSAs. For Boltz, OpenFold3, and Protenix, GreenFold-ColabFold partial is the default fallback: the ZIP includes a client-side helper that runs ColabFold/MMseqs2 only for missing chains and writes model-specific MSA files locally.
  • Non-protein entities are preserved in the rewritten input.
  • The ZIP includes the matching runner script package.
  • The ZIP includes RUN_COMMANDS.md with model-specific commands for the prepared input.
  • Use the report JSON to inspect selected UniProt IDs, scores, warnings, and fallback chains.
Open Documentation Local runner & automation

API Example


    


    
  


  
    

      
Start from an existing input

      

        The file is read into the entity list, which you can then edit like any other.
        Fields that cannot be represented as entities are reported rather than dropped silently.
      

    


    

      
      or drag a file here
      
      

    


    
or