Gemini Live Agent Challenge: Create Immersive AI Agents with Google Gemini Live

Google is launching a bold challenge to developers worldwide: abandon the simple text box to create immersive multimodal experiences. The Gemini Live Agent Challenge invites AI enthusiasts to push the boundaries of human-machine interaction by building agents that see, hear, speak, and create. With $80,000 in prizes and a deadline set for March 16, 2026, this hackathon organized by Google via Devpost represents a unique opportunity to shape the future of intelligent agents.

The chat box is officially too small for your ideas📦

The Gemini Live Agent Challenge is officially live! ⚡️@googleaidevs Multimodal Live API lets you build agents that listen, see, and react in the moment. What new experiences will you build?

🏆$80k in prizes
📅Deadline Mar… pic.twitter.com/77PqtA7Tze

— Devpost (@devpost) February 17, 2026

The Google Gemini Live Challenge: Beyond Text

The Gemini Live Agent Challenge poses a fundamental question: why limit AI to simple text exchanges when it can orchestrate complete experiences? Google is asking participants to develop a new next-generation AI agent that leverages multimodal inputs and outputs, building on Google’s Live API and video/image generation capabilities.

Projects must fall into one of these three strategic categories:

• Live Agents (Real-Time Agents): Create agents capable of natural real-time conversations, handling interruptions fluidly. Think instant translator, personalized tutor analyzing your homework through vision, or ultra-responsive voice customer assistant. Mandatory technologies: Gemini Live API or ADK (Agent Development Kit), hosted on Google Cloud.
• Creative Storyteller: Develop an agent that thinks like a creative director, weaving text, images, audio, and video into a single, coherent flow. Use Gemini Google’s native interleaved output capability to generate interactive books (text + illustrations), complete marketing assets (copy + visuals + video), or educational explanations (narration + diagrams).
• UI Navigator (Interface Navigator): Design an agent that becomes the user’s hands on screen. Your agent observes the browser or device display, interprets visual elements without necessarily relying on APIs or DOM access, and executes actions according to user intent. Mandatory technologies: Gemini AI multimodal to interpret screenshots/screen recordings, hosted on Google Cloud.

All projects must: use a Gemini model, be built with Google GenAI SDK OR ADK, and use at least one Google Cloud service.

Required Deliverables

Your submission must include:

• a complete text description (features, technologies, data sources, learnings),
• the URL of your public code repository (with deployment instructions in the README),
• proof of Google Cloud deployment (video recording showing console logs or link to code demonstrating use of GCP services),
• a clear architecture diagram (how Gemini connects to the backend, database, and frontend),
• a demonstration video of less than 4 minutes showing your multimodal/agentic features in actual action (no mockups) and pitching your project.

Practical Info & Gemini Google Hackathon Calendar

Format: Online hackathon, open to the public
Period: From March 16, 2026 at 8:00 PM EDT until March 16, 2026 at 5:00 PM PDT (deadline)
Time Zone: Deadline at 5:00 PM PDT (Pacific Time)

Key Dates & Prizes

• Submission Deadline: March 16, 2026 at 5:00 PM PDT
• Winner Announcement: Between April 22, 2026 and April 24, 2026

Prizes (Total $80,000):

• Grand Prize: $25,000 + $3,000 in Google Cloud credits + virtual coffee with Google team + social promotion + 2 Google Cloud Next 2026 tickets (April 22-24, 2026, value $2,299 each) + 2 travel stipends (max $3,000 each) + opportunity to demo at Google Cloud Next 2026
• Best Live Agent: $10,000 + $1,000 in Google Cloud credits + virtual coffee + social promotion + 2 Google Cloud Next 2026 tickets
• Best Creative Storyteller: $10,000 + $1,000 in Google Cloud credits + virtual coffee + social promotion + 2 Google Cloud Next 2026 tickets
• Best UI Navigator: $10,000 + $1,000 in Google Cloud credits + virtual coffee + social promotion + 2 Google Cloud Next 2026 tickets
• Best Multimodal Integration & UX: $5,000 + $500 in Google Cloud credits
• Best Technical Execution & Agent Architecture: $5,000 + $500 in Google Cloud credits
• Best Innovation & Thought Leadership: $5,000 + $500 in Google Cloud credits
• Honorable Mentions (5 winners): $2,000 each + $500 in Google Cloud credits

Who Can Participate in the Gemini Live Agent Challenge?

Eligibility: Participants of legal age of majority in their country of residence. Certain countries/territories are excluded (consult the full rules for the list).

Team Format: Solo and/or teams allowed, consult the rules

Technical Prerequisites: Knowledge of Google GenAI SDK or ADK, experience with Google Cloud, skills in developing multimodal AI agents

Participation Fee: Free

Level: Beginner to intermediate, official resources are provided to support all levels

Evaluation Criteria & Hackathon Rules

The jury will evaluate projects according to three main axes:

Innovation & Multimodal User Experience (40%): Does the project break the “text box” paradigm? Does the agent help “See, Hear, and Speak” fluidly? Does it have a distinct personality/voice? Is the experience “Live” and context-aware, or does it seem disjointed and sequential?

Technical Implementation & Agent Architecture (30%): Does the code effectively use the Google GenAI SDK or ADK? Is the backend robustly hosted on Google Cloud? Is the agent logic sound? Does it handle errors gracefully? Does the agent avoid hallucinations? Is there evidence of grounding?

Demo & Presentation (30%): Does the video clearly define the problem and solution? Is the architecture diagram clear? Is there visual proof of Cloud deployment? Does the video show the software actually working?

Important Rules

Projects must be new (developed for this hackathon), respect intellectual property rights, not use unauthorized data, and avoid any unethical or plagiarized content. Consult the full rules for all legal details and excluded countries.

Official Resources & Links

• Hackathon Registration: https://geminiliveagentchallenge.devpost.com/register
• Full Rules: https://geminiliveagentchallenge.devpost.com/rules
• Official Resources: https://geminiliveagentchallenge.devpost.com/resources
• Discussion Forum: https://geminiliveagentchallenge.devpost.com/forum_topics
• Contact: shawni@devpost.com

Why Participate in the Gemini Live Agent Challenge?

This hackathon represents much more than a simple competition. It’s an opportunity to shape the future of conversational AI alongside Google, one of the world leaders in artificial intelligence. You’ll gain access to cutting-edge technologies (Gemini Live API, ADK, Google Cloud), develop sought-after skills in multimodal agent development, and join a global community of 526+ already engaged participants.

The substantial prizes (up to $25,000 for the Grand Prize) include Google Cloud credits to continue your projects, networking opportunities with the Google team (virtual coffee), visibility through social promotion, and for the best, tickets to Google Cloud Next 2026 in Las Vegas with travel stipends, as well as an opportunity to present your solution to the Google ecosystem.

Beyond the rewards, you’ll build an impressive portfolio, discover the latest innovations in generative AI, and contribute to the evolution of human-machine interactions toward truly immersive experiences.