AttributeError: module 'google.cloud.aiplatform' has no attribute 'Agent' Despite Correct Installation

[this-dave]

Hi ADK Team,

We're encountering a persistent AttributeError when trying to access aip.Agent in our Python environment, even though all dependencies appear to be correctly installed.

Goal:
We are building an application that requires both local agent development (using adk web) and client-side interaction with a deployed Vertex AI Agent Engine agent.

Problem:
When attempting to import google.cloud.aiplatform as aip and use aip.Agent (e.g., in our Slack client app), we consistently receive the following error: AttributeError: module 'google.cloud.aiplatform' has no attribute 'Agent'.

Environment Details
OS: macOS (Apple Silicon, aarch64)

Python: 3.13

Package Manager: uv

What We've Tried
We have taken numerous steps to isolate the issue, but the error remains:

Clean Reinstalls: We have repeatedly deleted the virtual environment (.venv) and the lock file (uv.lock) and reinstalled everything from scratch using uv sync.

pyproject.toml Adjustments: We have tested various configurations, including a final version with a single, pinned dependency to avoid conflicts:

[project]
dependencies = [
    "google-cloud-aiplatform[adk,agent_engines]==1.56.0",
    # ... other dependencies like slack-bolt
]

Comprehensive Environment Diagnosis:

We confirmed that the PYTHONPATH environment variable is not set.

A diagnostic script verified that the correct Python interpreter from the .venv is being used and that the site-packages path is correct.

The script shows that version 1.56.0 of google-cloud-aiplatform is being loaded from the correct path within the .venv, but the Agent attribute is still missing from the imported module.

Isolated Test: We created a new, empty directory, set up a fresh virtual environment, and installed only the single package (uv pip install "google-cloud-aiplatform[adk,agent_engines]==1.56.0"). This isolated test also failed, proving the issue is not specific to our project's configuration.

Diagnostic Output:

--- Python Environment Diagnosis ---

[1] Python Executable:
    /path/to/project/.venv/bin/python

[4] Imported 'google.cloud.aiplatform' Library:
    Version: 1.56.0
    Location: /path/to/project/.venv/lib/python3.13/site-packages/google/cloud/aiplatform/__init__.py

--- Final Test ---

❌ FAILURE: module 'google.cloud.aiplatform' has no attribute 'Agent'

Our Question
This behavior strongly suggests that the [agent_engines] extra is not being installed or resolved correctly when using uv on a macOS/ARM system, even when the version number is pinned.

Are there any known incompatibilities between google-cloud-aiplatform[adk,agent_engines], Python 3.13, macOS (ARM), and/or uv?

What is the officially recommended pyproject.toml configuration and installation process for a project that requires both local adk web development and client-side aip.Agent usage?

Any help would be greatly appreciated, as this issue is currently blocking our development. Thank you!

[adk-bot]

Response from ADK Answering Agent (experimental, answer may be inaccurate)

TLDR: The AttributeError: module 'google.cloud.aiplatform' has no attribute 'Agent' error is likely because you should be using from google.adk.agents import Agent to define your agent, not aip.Agent.

---

Hi @this-dave,

I reviewed your issue and the ADK documentation. The problem seems to be in how you are trying to access the Agent class. Here's a breakdown and a suggested solution:

The Agent Class in ADK

Based on the ADK documentation, the Agent class for defining agents is part of the google.adk.agents module, not google.cloud.aiplatform. The google.cloud.aiplatform library is used for interacting with Vertex AI services, such as deploying to the Agent Engine, but not for defining the agent itself.

Recommended Code

To define your agent, you should use the following import:

from google.adk.agents import Agent

# Example agent definition
root_agent = Agent(
    name="my_agent",
    model="gemini-2.5-flash",
    description="A simple agent.",
    instruction="You are a helpful assistant.",
)

You can find more examples of how to define and use agents in the ADK documentation. [1]

pyproject.toml Configuration

Your pyproject.toml dependency on google-cloud-aiplatform[adk,agent_engines] is correct for projects that require both local ADK development and interaction with the Vertex AI Agent Engine. [1]

Recommendation

I recommend that you modify your code to import the Agent class from google.adk.agents and see if that resolves the AttributeError. If you are still facing issues, it might be helpful to create a minimal reproducible example based on the code snippets in the ADK documentation to ensure that your environment is set up correctly.

I hope this helps! Let me know if you have any other questions.

[1] gs://adk-knowledge-bucket-cloud-vi-test/adk-docs/docs/deploy/agent-engine.html