Error "Project/location and API key are mutually exclusive in the client initializer" for adk agent deployed using agent engine

[zmy1116]

Describe the bug
I notice that if GOOGLE_API_KEY is loaded in os.environ for adk agent, when it is deployed on agent engine, when running it we get error {'code': 498, 'message': 'Project/location and API key are mutually exclusive in the client initializer.'}

To Reproduce

• Create an agent with GOOGLE_API_KEY loaded

from google.adk.agents import Agent
from dotenv import load_dotenv
# Load environment variables
load_dotenv()
root_agent = Agent( ...)
...

• deploy the agent following using agent starter pack based on this tutorial https://google.github.io/adk-docs/deploy/agent-engine/

• Run it and get the error

LOCATION = "us-central1"
import json

import requests
import vertexai

client = vertexai.Client(
    location=LOCATION,
)
REASONING_ENGINE_ID = None

if REASONING_ENGINE_ID is None:
    try:
        with open("../deployment_metadata.json") as f:
            metadata = json.load(f)
            REASONING_ENGINE_ID = metadata.get("remote_agent_engine_id")
    except (FileNotFoundError, json.JSONDecodeError):
        pass
print(f"Using REASONING_ENGINE_ID: {REASONING_ENGINE_ID}")
# Get the existing agent engine
remote_agent_engine = client.agent_engines.get(name=REASONING_ENGINE_ID)

async for event in remote_agent_engine.async_stream_query(
    message="hi!", user_id="test"
):
    print(event)

Expected behavior
As the agent runs correctly when I launch the fastapi server myself (like using adk web), this error clearly is on the agent engine side. And I confirm if I don't load the GOOGLE_API_KEY, everything works fine

Why do I need to load GOOGLE_API_KEY: I have tools that involves gemini api calls. I guess I don't really have to get api key via the environment variable. So this bug is not blocking me on anything. But I think this error is still a bug.

Screenshots

Desktop (please complete the following information):

• OS: linux
• Python version(python -V): 3.10
• ADK version(pip show google-adk): 1.18

Model Information:

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

[zmy1116]

I want to provide additional problems I experienced:

So I thought the problem is just with GOOGLE_API_KEY in environment variables, the next thing is to load a renamed key. But that did not work. It looks like if the agent is on agent engine, I CANNOT launch genai client with google api key
Last few lines of the error log:

google.genai.errors.ClientError: 401 UNAUTHENTICATED. {'error': {'code': 401, 'message': 'API keys are not supported by this API. Expected OAuth2 access token or other authentication credentials that assert a principal. See https://cloud.google.com/docs/authentication', 'status': 'UNAUTHENTICATED', 'details': [{'@type': 'type.googleapis.com/google.rpc.ErrorInfo', 'reason': 'CREDENTIALS_MISSING', 'domain': 'googleapis.com', 'metadata': {'method': 'google.cloud.aiplatform.v1beta1.PredictionService.GenerateContent', 'service': 'aiplatform.googleapis.com'}}]}}

The next thing I did is to do not pass in api key when loading genai client. That worked for 2 turns of conversation and I got 429:

google.genai.errors.ClientError: 429 RESOURCE_EXHAUSTED. {'error': {'code': 429, 'message': 'Resource exhausted. Please try again later. Please refer to https://cloud.google.com/vertex-ai/generative-ai/docs/error-code-429 for more details.', 'status': 'RESOURCE_EXHAUSTED'}}

This does not happen when I use api key, does it mean this VERTEXAI has a separated rate limits that I need to apply for?