When using AgentTool or TeamTool with parallel tool calls enabled (the default), the second tool execution fails with "The team is already running, it cannot run again until it is stopped." This occurs because teams and agents maintain internal state that conflicts with concurrent execution.

The issue manifests when an LLM makes multiple tool calls simultaneously:

# This fails when the model makes parallel tool calls
main_agent = AssistantAgent(
    name="main_agent",
    model_client=model_client,  # parallel_tool_calls=True by default
    tools=[team_tool],
)

# The LLM might generate:
# [FunctionCall(name='writing_team', ...), FunctionCall(name='writing_team', ...)]
# Second call fails: "The team is already running"

Solution: Users must disable parallel tool calls when using these tools:

# Correct usage - disable parallel tool calls
model_client = OpenAIChatCompletionClient(model="gpt-4", parallel_tool_calls=False)
main_agent = AssistantAgent(
    name="main_agent", 
    model_client=model_client,
    tools=[team_tool],
)

This PR adds comprehensive documentation warnings:

1. Tutorial documentation (agents.ipynb): Added prominent warning in the Parallel Tool Calls section explaining the limitation and solution
2. TeamTool docstring: Added .. important:: directive with clear explanation and updated example showing proper usage
3. AgentTool docstring: Added matching warning and example demonstrating parallel_tool_calls=False

All examples now show the correct pattern of creating separate model clients with parallel tool calls disabled for the main agent while allowing the wrapped agents/teams to use their own model clients normally.

Fixes #6882.

💬 Share your feedback on Copilot coding agent for the chance to win a $200 gift card! Click here to start the survey.