diff --git a/src/genomic_features/ensembl/ensembldb.py b/src/genomic_features/ensembl/ensembldb.py
index a731470..729dba8 100644
--- a/src/genomic_features/ensembl/ensembldb.py
+++ b/src/genomic_features/ensembl/ensembldb.py
@@ -7,6 +7,7 @@
 from typing import Final, Literal
 
 import ibis
+import numpy as np
 import requests
 from ibis import deferred
 from ibis.expr.types import Table as IbisTable
@@ -269,6 +270,56 @@ def chromosomes(self) -> DataFrame:
         """
         return self.db.table("chromosome").execute()
 
+    def promoters(
+        self,
+        filter: _filters.AbstractFilterExpr = filters.EmptyFilter(),
+        upstream: int = 2000,
+        downstream: int = 200,
+        canonical_transcripts: bool = False,
+    ) -> DataFrame:
+        """Get promoter annotations.
+
+        Parameters
+        ----------
+        filter
+            Filter expression to apply to the genes table.
+        upstream
+            Number of base pairs upstream of the TSS (default: 2000).
+        downstream
+            Number of base pairs downstream of the TSS (default: 200).
+        canonical_transcripts
+            If True, return only canonical transcript for each gene (default: False).
+
+        Returns
+        -------
+        DataFrame
+            A table of promoter annotations.
+        """
+        # TODO: change to get transcript table with gene level columns
+        # something like:
+        # tx_table = self.transcripts(cols = set(cols + ['seq_strand', 'seq_name', 'tx_is_canonical']), filter = filter)
+        tx_table = self.genes(filter=filter)
+
+        # Get promoter region based on strand
+        # strand = 1         |>>>>>>>>>>>>>>|
+        # strand = -1                         |<<<<<<<<<<<<<<|
+        # Tx SS:             *                               *
+        # Promoter       ------                             ------
+        tx_ss = np.where(
+            tx_table["seq_strand"] == 1,
+            tx_table["gene_seq_start"],
+            tx_table["gene_seq_end"],
+        )
+        tx_table["promoter_seq_start"] = np.where(
+            tx_table["seq_strand"] == 1, tx_ss - upstream, tx_ss - downstream
+        )
+        tx_table["promoter_seq_end"] = np.where(
+            tx_table["seq_strand"] == 1, tx_ss + downstream, tx_ss + upstream
+        )
+        # if canonical_transcripts:
+        #     tx_table = tx_table[tx_table["tx_is_canonical"] == 1]
+        return tx_table
+
     def _build_query(
         self,
         table: Literal["gene", "tx", "exon"],
diff --git a/tests/test_filters.py b/tests/test_filters.py
index 33d5756..4bbee4d 100644
--- a/tests/test_filters.py
+++ b/tests/test_filters.py
@@ -205,3 +205,22 @@ def test_seqs_as_int(hsapiens108):
 
     pd.testing.assert_frame_equal(result_w_ints, result_w_strs)
     pd.testing.assert_frame_equal(result_w_ints, result_w_mixed)
+
+
+def test_promoters(hsapiens108):
+    promoters = hsapiens108.promoters()
+    assert isinstance(promoters, pd.DataFrame)
+    promoters = hsapiens108.promoters(upstream=100, downstream=100)
+    assert ((promoters.promoter_seq_end - promoters.promoter_seq_start) == 200).all()
+    promoters = hsapiens108.promoters(upstream=1000, downstream=100)
+    assert ((promoters.promoter_seq_end - promoters.promoter_seq_start) == 1100).all()
+    # Test strandedness
+    promoters = hsapiens108.promoters(upstream=1000, downstream=100)
+    assert (
+        promoters[promoters.seq_strand == -1].promoter_seq_start
+        == promoters[promoters.seq_strand == -1].gene_seq_end - 100
+    ).all()
+    assert (
+        promoters[promoters.seq_strand == 1].promoter_seq_start
+        == promoters[promoters.seq_strand == 1].gene_seq_start - 1000
+    ).all()