diff --git a/surfsense_backend/alembic/versions/124_add_token_usage_table.py b/surfsense_backend/alembic/versions/124_add_token_usage_table.py
new file mode 100644
index 000000000..2c88e1f51
--- /dev/null
+++ b/surfsense_backend/alembic/versions/124_add_token_usage_table.py
@@ -0,0 +1,81 @@
+"""124_add_token_usage_table
+
+Revision ID: 124
+Revises: 123
+Create Date: 2026-04-14
+
+Adds token_usage table for tracking LLM token consumption per message.
+Supports future extension via usage_type for indexing, image gen, etc.
+"""
+
+from __future__ import annotations
+
+from collections.abc import Sequence
+
+import sqlalchemy as sa
+from sqlalchemy.dialects.postgresql import JSONB, UUID
+
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision: str = "124"
+down_revision: str | None = "123"
+branch_labels: str | Sequence[str] | None = None
+depends_on: str | Sequence[str] | None = None
+
+
+def upgrade() -> None:
+    conn = op.get_bind()
+    if sa.inspect(conn).has_table("token_usage"):
+        return
+
+    op.create_table(
+        "token_usage",
+        sa.Column("id", sa.Integer(), primary_key=True, autoincrement=True),
+        sa.Column("prompt_tokens", sa.Integer(), nullable=False, server_default="0"),
+        sa.Column("completion_tokens", sa.Integer(), nullable=False, server_default="0"),
+        sa.Column("total_tokens", sa.Integer(), nullable=False, server_default="0"),
+        sa.Column("model_breakdown", JSONB, nullable=True),
+        sa.Column("call_details", JSONB, nullable=True),
+        sa.Column("usage_type", sa.String(50), nullable=False, server_default="chat"),
+        sa.Column(
+            "thread_id",
+            sa.Integer(),
+            sa.ForeignKey("new_chat_threads.id", ondelete="CASCADE"),
+            nullable=True,
+        ),
+        sa.Column(
+            "message_id",
+            sa.Integer(),
+            sa.ForeignKey("new_chat_messages.id", ondelete="SET NULL"),
+            nullable=True,
+        ),
+        sa.Column(
+            "search_space_id",
+            sa.Integer(),
+            sa.ForeignKey("searchspaces.id", ondelete="CASCADE"),
+            nullable=False,
+        ),
+        sa.Column(
+            "user_id",
+            UUID(as_uuid=True),
+            sa.ForeignKey("user.id", ondelete="CASCADE"),
+            nullable=False,
+        ),
+        sa.Column(
+            "created_at",
+            sa.TIMESTAMP(timezone=True),
+            nullable=False,
+            server_default=sa.func.now(),
+        ),
+    )
+
+    op.create_index("ix_token_usage_thread_id", "token_usage", ["thread_id"])
+    op.create_index("ix_token_usage_message_id", "token_usage", ["message_id"])
+    op.create_index("ix_token_usage_search_space_id", "token_usage", ["search_space_id"])
+    op.create_index("ix_token_usage_user_id", "token_usage", ["user_id"])
+    op.create_index("ix_token_usage_usage_type", "token_usage", ["usage_type"])
+
+
+def downgrade() -> None:
+    op.drop_table("token_usage")
diff --git a/surfsense_backend/app/db.py b/surfsense_backend/app/db.py
index e69d28ac2..b9fbe8845 100644
--- a/surfsense_backend/app/db.py
+++ b/surfsense_backend/app/db.py
@@ -647,6 +647,11 @@ class NewChatThread(BaseModel, TimestampMixin):
         cascade="all, delete-orphan",
         foreign_keys="[PublicChatSnapshot.thread_id]",
     )
+    token_usages = relationship(
+        "TokenUsage",
+        back_populates="thread",
+        cascade="all, delete-orphan",
+    )
 
 
 class NewChatMessage(BaseModel, TimestampMixin):
@@ -685,6 +690,63 @@ class NewChatMessage(BaseModel, TimestampMixin):
         back_populates="message",
         cascade="all, delete-orphan",
     )
+    token_usage = relationship(
+        "TokenUsage",
+        back_populates="message",
+        uselist=False,
+        cascade="all, delete-orphan",
+    )
+
+
+class TokenUsage(BaseModel, TimestampMixin):
+    """
+    Tracks LLM token consumption per assistant turn.
+
+    One row per usage event. For chat, linked to a specific message via message_id.
+    The usage_type column enables future extension to track non-chat usage
+    (indexing, image generation, podcasts, etc.) without schema changes.
+    """
+
+    __tablename__ = "token_usage"
+
+    prompt_tokens = Column(Integer, nullable=False, default=0)
+    completion_tokens = Column(Integer, nullable=False, default=0)
+    total_tokens = Column(Integer, nullable=False, default=0)
+    model_breakdown = Column(JSONB, nullable=True)
+    call_details = Column(JSONB, nullable=True)
+
+    usage_type = Column(String(50), nullable=False, default="chat", index=True)
+
+    thread_id = Column(
+        Integer,
+        ForeignKey("new_chat_threads.id", ondelete="CASCADE"),
+        nullable=True,
+        index=True,
+    )
+    message_id = Column(
+        Integer,
+        ForeignKey("new_chat_messages.id", ondelete="SET NULL"),
+        nullable=True,
+        index=True,
+    )
+    search_space_id = Column(
+        Integer,
+        ForeignKey("searchspaces.id", ondelete="CASCADE"),
+        nullable=False,
+        index=True,
+    )
+    user_id = Column(
+        UUID(as_uuid=True),
+        ForeignKey("user.id", ondelete="CASCADE"),
+        nullable=False,
+        index=True,
+    )
+
+    # Relationships
+    thread = relationship("NewChatThread", back_populates="token_usages")
+    message = relationship("NewChatMessage", back_populates="token_usage")
+    search_space = relationship("SearchSpace")
+    user = relationship("User")
 
 
 class PublicChatSnapshot(BaseModel, TimestampMixin):