Inspiration
Marketing budgets are often managed reactively. By the time a human notices a campaign is failing, money has already been wasted. We wanted to build an agent that acts with the speed of an algorithm but the strategic caution of a veteran marketer.
What it does
Alpha-Agent operates on a continuous Sense → Decide → Act → Observe → Learn loop:
Autonomous Reallocation: It identifies "Winners" and "Losers" and shifts budgets instantly to capture peak ROI windows.
External Intelligence: It uses You.com to verify if a CPA spike is due to internal campaign issues or external market volatility (like a competitor price surge).
Self-Regulating Autonomy: The agent tracks its own success rate. If its decisions lead to performance decay, it automatically reduces its "Aggressiveness" and can trigger a System Freeze via Plivo SMS for human intervention.
How we built it
Frontend: Built with Lovable, focusing on "Visual Agency" so users can see the "money flow" between campaigns.
Backend & Logic: Developed in Cursor using a deterministic policy engine to ensure spend decisions are predictable and safe.
Integrations: *
You.com: Provides real-time web search to ground the agent's decisions in market truth.
Deployment: Hosted on Render for persistent, 24/7 autonomous operation.
Challenges we ran into
The biggest challenge was Explainable Agency. It’s easy to move numbers; it’s hard to explain why they were moved in a way a human can trust. We solved this by building a detailed Intent Log that translates math into natural language strategy.
Accomplishments that we're proud of
Successfully stitching together Composio and You.com to create an agent that isn't just a "black box," but a market-aware system that can justify its actions with live web data.
What we learned
We learned that the hardest part of AI agents isn't the "AI"—it's the Human-in-the-Loop design. Trust is built when the agent knows its limits and asks for help before things go wrong.
What's next for Alpha-Agent
We plan to expand the "Sense" layer to include social media sentiment (X/Twitter) to predict viral trends before they hit the ad auctions, allowing the agent to be even more proactive.
Log in or sign up for Devpost to join the conversation.