Thanks — I’m responding to your question about forcing an agent to call one and only one tool at a time. Yes — you can enforce sequential (one-at-a-time) tool calls by disabling parallel tool calls. Set the LLM configuration parameter

parallel_tool_calls

to

False

. When

parallel_tool_calls

is disabled, the agent makes tool calls sequentially rather than concurrently. Example (Python):

session = AgentSession(
    llm=anthropic.LLM(
        model="claude-3-5-sonnet-20241022",
        parallel_tool_calls=False,  # Ensures one tool call at a time
    ),
    # ... other options
)

By default,

parallel_tool_calls

is not enabled, so unless you explicitly set it to

True

, agents will typically call tools one at a time. If you’d like, tell me which LLM provider or SDK you’re using and I can give a tailored code snippet or confirm the exact setting name for that provider. Sources: Anthropic Claude LLM plugin guide | LiveKit Documentation | Fireworks AI LLM plugin guide | LiveKit Documentation | DeepSeek LLM | LiveKit Documentation | Telnyx LLM plugin guide | LiveKit Documentation | Cerebras LLM plugin guide | LiveKit Documentation