AEON — Agent Procurement Control Plane
Inspiration
As AI agents evolve from assistants into autonomous decision-makers, we asked: What happens when an AI can spend money on behalf of an organization? Autonomy without accountability creates financial risks. We envisioned AEON as the governance layer that makes autonomous commerce trustworthy.
What it does
AEON enables AI agents to initiate procurement while keeping organizations in control. It authenticates agents, collects supplier quotes, evaluates options through ML-assisted recommendations, and applies deterministic policies to spending limits, supplier permissions, and budgets. Requests requiring additional authorization go to human reviewers. Approved purchases proceed through PayPal, with payment reconciliation and an auditable transaction history.
How we built it
We built the backend using Python, FastAPI, SQLAlchemy, PostgreSQL, and Alembic, with a React, TypeScript, Vite, and Tailwind CSS frontend. Modular services handle authentication, procurement, policy evaluation, budget management, approvals, and PayPal integration. A recommendation adapter separates ML-assisted evaluation from authorization decisions.
Challenges we ran into
Integrating frontend and backend contracts, configuring database environments, and testing PayPal Sandbox workflows were challenging. Distinguishing buyer approval from successful payment capture was particularly important. We also had to consider concurrent spending, duplicate payment requests, and reliable webhook processing.
Accomplishments that we're proud of
We established a complete procurement architecture combining agent authentication, policy enforcement, human oversight, payment orchestration, and auditability. Most importantly, AEON separates intelligent recommendations from financial authorization.
What we learned
We learned that AI recommendations should never replace deterministic financial controls. Secure automation requires transaction integrity, idempotency, transparent decisions, and meaningful human oversight.
What's next for AEON
We plan to strengthen payment reconciliation, webhook reliability, and automated testing, expand real supplier integrations, improve procurement analytics, and prepare the platform for production deployment. We built this project in the strict timeframe of the hackathon focusing on our central principles.
ML recommends. Policies govern. Humans oversee. AEON authorizes.
Built With
- alembic
- css3
- fastapi
- html5
- javascript
- jwt-authentication
- machine-learning
- orders
- paypal
- paypal-orders-api
- postgresql
- pydantic
- pytest
- python
- react
- render
- restapi
- sha-256
- sqlalchemy
- supabase
- tailwind
- typescript
- vercel
- vite

