maximumRemoteCalls does not limit repeated tool calls when using Agent Config with functionCallingConfig.mode: ANY

[hapiiiiGit]

🔴 Required Information

Describe the Bug:
When using ADK’s Agent Config mode with Gemini function calling configured as mode: ANY, I expected automaticFunctionCalling.maximumRemoteCalls to limit the number of tool calls made during a task.
However, in my run, even though maximumRemoteCalls was set to 4, the agent repeatedly called the same tool far more than 4 times.

Steps to Reproduce:

1. root_agent.yaml:

agent_class: LlmAgent

name: my_agent
model: gemini-2.5-pro
description: Agent for HumanEval-style Python code generation experiments.

instruction: |
  You are a Python coding assistant. Your task is to complete the following Python code from the HumanEval benchmark.  Use tools when they help, and give concise, evidence-aware answers.
tools:
  - name: my_agent_google.tools.submit_answer
  - name: my_agent_google.tools.python_executor
 generate_content_config:
  temperature: 0.0
  responseMimeType: text/plain
  toolConfig:
    functionCallingConfig:
      mode: ANY
  automaticFunctionCalling:
    disable: false
    maximumRemoteCalls: 4
    ignoreCallHistory: true

2. input:

adk run my_agent
Please complete the code: from typing import List; def has_close_elements(numbers: List[float], threshold: float) -> bool: """Check if in given list of numbers, are any two numbers closer to each other than given threshold."""

Observed Behavior:
The agent called submit_answer once successfully.
But after that, the model continued calling submit_answer repeatedly. Each subsequent call returned:

{
  "ok": false,
  "error": "submit_answer has already been called",
  "system_directive": "submit_answer may only be called once and must be the final tool call. Do not call submit_answer or any other tool again."
}

This repeated far more than 4 times, even though maximumRemoteCalls was set to 4.

Environment Details:

• ADK Library Version (pip show google-adk): adk 1.33.0
• Desktop OS:** [e.g., macOS, Linux, Windows] Windows
• Python Version (python -V): Python 3.11.15

Model Information:

• Are you using LiteLLM: No
• Which model is being used: gemini-2.5-pro

---

🟡 Optional Information

Providing this information greatly speeds up the resolution process.

log

mode-any-log.txt

Screenshots / Video:
If applicable, add screenshots or screen recordings to help explain
your problem.
N/A

Additional Context:
Add any other context about the problem here.
N/A

Minimal Reproduction Code:
Please provide a code snippet or a link to a Gist/repo that isolates the issue.

// Code snippet here

N/A

How often has this issue occurred?:

• Always (100%)

[sanketpatil06]

Hi @hapiiiiGit ,Thanks for the detailed report. We were able to reproduce this and observed same behaviour as you mentioned.

• ANY mode strictly forces the model to output a tool call, removing its ability to return standard text.
• AUTO mode allows the model to decide whether to call a tool or conclude with a text response.

When ADK hits the maximumRemoteCalls limit, it stops executing tools. However, because mode: ANY is still active, the model is physically forbidden from returning a final text response, causing a deadlock loop.
Workaround worth trying: Use mode: AUTO combined with strict prompt instructions. This ensures the model uses tools but can still gracefully exit the loop.

instruction: |
  You are a Python coding assistant. 
  IMPORTANT: You MUST use the `python_executor` tool before providing your final answer!

generate_content_config:
  tool_config:
    function_calling_config:
      mode: AUTO  # Downgrade to AUTO to allow text generation
  automatic_function_calling:
    maximum_remote_calls: 4

Could you please give it a try and let us know if this works for you? Thanks.

[hapiiiiGit]

Hi @sanketpatil06, thanks for the explanation and workaround.
I tried switching to AUTO mode, and it does allow the LLM to output text and exit the loop.
Does this mean maximum_remote_calls is only effective in AUTO mode?