Clarix- Google Ads Marketing Assistant

• 

  Generates smart marketing content using multi-agent AI

Inspiration

Clarix was inspired by my digital marketer friend for writing effective content while he uses Google Ads and works as a personal assistant that will understand him and his business. And an AI agent that will understand the audience and write content according to that customer and the right tone and formats for the right customer.

What it does

Clarix is a multi-agent that:

1. Identifies the target audience
   
2. Writes a social media ad tailored to that audience
3. Generates promotional email
4. writes content in a specific tone (e.g., formal, playful, etc.)

How we built it

Built Clarix using Google Agent Development Kit (ADK) and Python Created four LlmAgents:

1. audience_agent
2. ad_writer_agent
3. email_writer_agent
4. tone_optimizer_agent

The root agent invokes the sub-agents. Used .env for secure API key handling Tested using ADK Web

Challenges we ran into

1.Resolved import issues with google.aadk.runtime.run_agnet and switched to .invoke()

1. Ensuring ADK Web could detect the root_agent correctly.
2. Managed modularity without introducing circular imports.
3. Balanced detailed output while keeping each agent’s task focused

Accomplishments that we're proud of

1. Design clean, scalable code with meaningful agent separation.
2. Successfully built first multi-agent workflow using ADK.
3. Made a system that is functional and fast.

What we learned

1. learned how to orchestrate and architect agent collaboration using ADK
2. Practical use of tools and environment management.
3. How to debug and structure projects with ADK Web and Cloud.
4. Thinking like a product designer, not just a coder.

What's next for Clarix,- Google Ads Marketing Assistant

1. Voice integration will be helpful for those who need hands-free control.
2. Deeper market understanding that will learn from the user and understand the person's business and will provide effective decisions.
3. Analytics Agent that will analyze engagement data from previous posts/emails and suggest the best time, tone, and platform for the next campaign.

Built With

• .invoke()
• adk
• agenttool
• gemini2.5flash
• llmagent
• python3.10