diff --git a/src/backend/base/langflow/services/tracing/service.py b/src/backend/base/langflow/services/tracing/service.py
index 6eef40cf3..7b21c5988 100644
--- a/src/backend/base/langflow/services/tracing/service.py
+++ b/src/backend/base/langflow/services/tracing/service.py
@@ -1,5 +1,12 @@
-from typing import TYPE_CHECKING
+import os
+import traceback
+from contextlib import contextmanager
+from typing import TYPE_CHECKING, Any, Dict
 
+from langchain.callbacks.tracers.langchain import wait_for_all_tracers
+from loguru import logger
+
+from langflow.schema.data import Data
 from langflow.services.base import Service
 
 if TYPE_CHECKING:
@@ -13,3 +20,163 @@ class TracingService(Service):
     def __init__(self, settings_service: "SettingsService", monitor_service: "MonitorService"):
         self.settings_service = settings_service
         self.monitor_service = monitor_service
+        self.inputs = {}
+        self.outputs = {}
+        self.outputs_metadata = {}
+        self.run_name = None
+        self.run_id = None
+        self.project_name = None
+        self._tracers: dict[str, LangSmithTracer] = {}
+
+    def _reset_io(self):
+        self.inputs = {}
+        self.outputs = {}
+
+    def initialize_tracers(self):
+        self._initialize_langsmith_tracer()
+
+    def _initialize_langsmith_tracer(self):
+        project_name = os.getenv("LANGCHAIN_PROJECT", "Langflow")
+        self.project_name = project_name
+        self._tracers["langsmith"] = LangSmithTracer(
+            trace_name=self.run_name,
+            trace_type="chain",
+            project_name=self.project_name,
+            trace_id=self.run_id,
+        )
+
+    def set_run_name(self, name: str):
+        self.run_name = name
+
+    def set_run_id(self, run_id: str):
+        self.run_id = run_id
+
+    def _start_traces(self, trace_name: str, trace_type: str, inputs: Dict[str, Any], metadata: Dict[str, Any] = None):
+        for tracer in self._tracers.values():
+            if not tracer.ready:
+                continue
+            tracer.add_trace(trace_name, trace_type, inputs, metadata)
+
+    def _end_traces(self, trace_name: str, error: str | None = None):
+        for tracer in self._tracers.values():
+            if not tracer.ready:
+                continue
+            tracer.end_trace(trace_name=trace_name, outputs=self.outputs, error=error)
+
+    def _end_all_traces(self, outputs: dict[str, Any]):
+        for tracer in self._tracers.values():
+            if not tracer.ready:
+                continue
+            tracer.end(outputs=outputs)
+
+    def end(self, outputs: dict[str, Any] | None = None):
+        self._end_all_traces(outputs)
+        self._reset_io()
+
+    @contextmanager
+    def trace_context(
+        self, trace_name: str, trace_type: str, inputs: Dict[str, Any] = None, metadata: Dict[str, Any] = None
+    ):
+        self._start_traces(trace_name, trace_type, inputs, metadata)
+        try:
+            yield self
+        except Exception as e:
+            tb = traceback.format_exc()
+            error_message = f"{e.__class__.__name__}: {e}\n\n{tb}"
+            self._end_traces(trace_name, error_message)
+            raise e
+        finally:
+            self._end_traces(trace_name, None)
+            self._reset_io()
+
+    def set_outputs(self, outputs: Dict[str, Any], output_metadata: Dict[str, Any] = None):
+        self.outputs = outputs
+        self.outputs_metadata = output_metadata
+
+
+class LangSmithTracer:
+    def __init__(self, trace_name: str, trace_type: str, project_name: str, trace_id: str):
+        from langsmith.run_trees import RunTree
+
+        self.trace_name = trace_name
+        self.trace_type = trace_type
+        self.project_name = project_name
+        self.trace_id = trace_id
+        self._run_tree = RunTree(
+            project_name=self.project_name,
+            name=self.trace_name,
+            run_type=self.trace_type,
+            id=self.trace_id,
+        )
+        self._children: dict[str, RunTree] = {}
+        self._ready = self.setup_langsmith()
+
+    @property
+    def ready(self):
+        return self._ready
+
+    def setup_langsmith(self):
+        try:
+            from langsmith import Client
+
+            self._client = Client()
+        except ImportError:
+            logger.error("Could not import langsmith. Please install it with `pip install langsmith`.")
+            return False
+        os.environ["LANGCHAIN_TRACING_V2"] = "true"
+        return True
+
+    def add_trace(self, trace_name: str, trace_type: str, inputs: Dict[str, Any], metadata: Dict[str, Any] = None):
+        if not self._ready:
+            return
+        inputs = self._convert_to_langchain_types(inputs)
+        child = self._run_tree.create_child(
+            name=trace_name,
+            run_type=trace_type,
+            inputs=inputs,
+        )
+        if metadata:
+            child.add_metadata(metadata)
+        self._children[trace_name] = child
+
+    def _convert_to_langchain_types(self, io_dict: Dict[str, Any]):
+        from langflow.schema.message import Message
+
+        _converted = {}
+        for key, value in io_dict.items():
+            if isinstance(value, dict):
+                _converted[key] = self._convert_to_langchain_types(value)
+            elif isinstance(value, list):
+                _converted[key] = [self._convert_to_langchain_types(v) for v in value]
+            elif isinstance(value, Message):
+                if value.sender:
+                    _converted[key] = value.to_lc_message()
+                elif "prompt" in value:
+                    _converted[key] = value.load_lc_prompt()
+                else:
+                    _converted[key] = value.to_lc_document()
+            elif isinstance(value, Data):
+                _converted[key] = value.to_lc_document()
+            else:
+                _converted[key] = value
+        return _converted
+
+    def end_trace(self, trace_name: str, outputs: Dict[str, Any] = None, error: str = None):
+        child = self._children[trace_name]
+        raw_outputs = {}
+        processed_outputs = {}
+        if outputs:
+            raw_outputs = outputs
+            processed_outputs = self._convert_to_langchain_types(outputs)
+        child.add_metadata({"outputs": raw_outputs})
+        child.end(outputs=processed_outputs, error=error)
+        if error:
+            child.patch()
+        else:
+            child.post()
+
+    def end(self, outputs: Dict[str, Any]):
+        self._run_tree.end(outputs=outputs)
+        self._run_tree.post()
+        wait_for_all_tracers()
+        self._run_link = self._run_tree.get_url()