TypeError: 'ToolNode' object is not iterable in patch.py when using langgraph-supervisor with create_react_agent

[himadri-bhattacharjee]

Component: Langchain Instrumentation

Description:

opentelemetry-instrumentation-langchain v0.53.x (also present in v0.57.0.) crashes when langgraph-supervisor's create_supervisor() passes a ToolNode object to
create_react_agent. The instrumentation's monkey-patch in patch.py , assumes tools is always an iterable list, but ToolNode is not
iterable.

Steps to Reproduce:

1. Install opentelemetry-instrumentation-langchain>=0.53.3 with langgraph-supervisor
2. Create a supervisor using create_supervisor() which internally calls create_react_agent with a ToolNode object
3. Application crashes on startup

Error:
File "/usr/local/lib/python3.13/site-packages/opentelemetry/instrumentation/langchain/patch.py", line 489, in wrapper
for tool in tools:
^^^^^
TypeError: 'ToolNode' object is not iterable

Root Cause:

In patch.py lines 484-492:
python
tools = kwargs.get("tools")
if tools is None and len(args) > 1:
tools = args[1]
if tools:
tool_definitions = []
for tool in tools: # <-- ToolNode is not iterable
tool_def = _extract_tool_definition(tool)

ToolNode is a LangGraph class that wraps tools but is not itself iterable. It exposes tools via tools_by_name attribute.

Suggested Fix:

python
if tools:
tool_definitions = []
_tools_iter = tools.tools_by_name.values() if hasattr(tools, 'tools_by_name') else tools
for tool in _tools_iter:

Workaround: Pin to opentelemetry-instrumentation-langchain==0.52.6

Environment:

• opentelemetry-instrumentation-langchain: 0.53.3 (upto v0.57.0)
• langgraph-prebuilt: 1.0.8
• langgraph-supervisor: latest
• Python: 3.13

[himadri-bhattacharjee]

same issue reported , #3841

[JiwaniZakir]

The root cause is in patch.py around line 489, where the instrumentation assumes tools is always a plain iterable (list/tuple), but ToolNode is a LangGraph graph node that wraps tools internally — it doesn't implement __iter__. The TypeError: 'ToolNode' object is not iterable confirms exactly this: Python's for loop calls iter() on the object, which fails because ToolNode has no __iter__ method. The suggested fix using tools.tools_by_name.values() is the right approach, since ToolNode stores its wrapped tools in that dict. It may also be worth guarding against other non-iterable tool container types that could appear in future LangGraph versions by adding a broader isinstance check or a try/except TypeError fallback rather than relying solely on the tools_by_name attribute name.