AgentOS Pro

Inspiration

For small and medium-sized businesses (SMBs), managing daily operations, strategic forecasting, sales pipelines, and marketing campaigns is a constant struggle of context-switching. Traditional SaaS tools are static and fragmented. We asked ourselves: What if a business could run on an autonomous operating system powered by specialized AI agents that learn, collaborate, and predict the future? That inspired us to build AgentOS Pro—an autonomous AI-native business operating system that empowers SMBs to automate complex workflows and predict demand using advanced machine learning.

What it does

AgentOS Pro is a multi-tenant platform that deploys a team of 8 specialized, autonomous AI agents to run and optimize an SMB:

1. Clark (Command Center): Coordinates the agent team and provides executive daily briefings.
2. Lionel (Finance Agent): Automates cash flow tracking, expense analysis, and invoice reconciliation.
3. Oracle (Forecast Agent): Powered by Google’s TimesFM model, it predicts sales, revenue, and inventory needs.
4. Hunter (Sales Agent): Automates lead qualification, CRM updates, and B2B sales outreach.
5. Jules (Content Strategist): Crafts marketing calendars, social media copy, and automated content pipelines.
6. Milo (Competitive Intel): Monitors competitors, analyzes pricing, and generates positioning battle cards.
7. Chloe (Market Analyst): Analyzes customer reviews and feedback to build buyer personas.
8. Oliver (Operations Agent): Optimizes supply chain logistics and inventory reorder points.

How we built it

We architected AgentOS Pro with a cutting-edge AI stack:

• AI Orchestration: Built using the Google Antigravity SDK, utilizing its advanced agent capabilities, parent-child delegation (subagents), structured outputs, and periodic cron-like triggers.
• Predictive Core: Integrated Google Research's TimesFM 2.0 model deployed via Vertex AI to perform zero-shot time-series forecasting.
• Backend: Powered by FastAPI (Python) for async performance and RESTful API endpoints.
• Frontend: A sleek, responsive dashboard built with Next.js and styled with modern CSS, featuring glassmorphism and real-time visualization of agent activity.
• Infrastructure: Hosted securely on Google Cloud (using Cloud Run, Cloud SQL, and Cloud Storage) for robust scaling.

Challenges we ran into

• Multi-Agent Orchestration: Coordinating 8 agents so they can share context without getting stuck in infinite loops. We solved this by using the Antigravity SDK's robust policy system and a centralized event bus.
• Time-Series Forecasting: Mapping sparse SMB data into TimesFM input formats. We built data-normalization pipelines that preprocess Shopify and CSV data before feeding them to the forecasting engine.
• Real-time UI: Keeping the user dashboard updated with real-time agent thoughts and actions. We achieved this through Server-Sent Events (SSE) and optimized state updates.

Accomplishments that we're proud of

• Building a fully functional multi-tenant SaaS from scratch during the hackathon period.
• Integrating Google's TimesFM model to provide SMBs with enterprise-grade forecasting.
• Creating a beautiful, unified dashboard where users can watch their autonomous agent team work in real time.
• Generating real revenue from pilot SMB customers within the 90-day hackathon window.

What we learned

We learned the power of the Google Antigravity SDK for rapid agent development and how zero-shot time-series models like TimesFM can outperform traditional statistical models on small business datasets. We also learned how to balance autonomy with human-in-the-loop controls.

What's next for AgentOS Pro

We plan to expand our integrations to support more e-commerce platforms (like WooCommerce and Amazon), add custom tool creation for users, and implement deeper voice capabilities for agent interactions.

Built With

• css3
• fastapi
• gemini-api
• google-antigravity-sdk
• google-cloud
• html5
• javascript
• next.js
• python
• timesfm
• vertex-ai