diff --git a/docs/source/input_format.md b/docs/source/input_format.md
index 98f0a686..acb02525 100644
--- a/docs/source/input_format.md
+++ b/docs/source/input_format.md
@@ -55,6 +55,7 @@ All chains must define a unique ```chain_ids``` field and appropriate sequence o
   {
     "molecule_type": "protein",
     "chain_ids": "A",
+    "description": "Optional metadata example",
     "sequence": "PVLSCGEWQCL",
     "use_msas": true,
     "use_main_msas": true,
@@ -72,6 +73,9 @@ All chains must define a unique ```chain_ids``` field and appropriate sequence o
   - `chain_ids` *(str | list[str], required)*
     - One or more identifiers for this chain. Used to map sequences to structure outputs.
 
+  - `description` *(str | None, optional, default = null)
+    - Optional metadata to provide for each chain.
+
   - `sequence` *(str, required)*
     - Amino acid sequence (1-letter codes), supporting standard residues, X (unknown), and U (selenocysteine).
 
diff --git a/openfold3/projects/of3_all_atom/config/inference_query_format.py b/openfold3/projects/of3_all_atom/config/inference_query_format.py
index db9e9626..0fa7d209 100644
--- a/openfold3/projects/of3_all_atom/config/inference_query_format.py
+++ b/openfold3/projects/of3_all_atom/config/inference_query_format.py
@@ -49,6 +49,7 @@ class Chain(BaseModel):
     }
     molecule_type: Annotated[MoleculeType, BeforeValidator(_convert_molecule_type)]
     chain_ids: Annotated[list[str], BeforeValidator(_ensure_list)]
+    description: str | None = None
     sequence: str | None = None
     non_canonical_residues: (
         Annotated[dict[int, str], BeforeValidator(_cast_keys_to_int)] | None