-
-
INTENTRA Command Center — AI-powered purchasing, human-controlled spending, and real-time financial oversight.
-
INTENTRA AI Purchasing Assistant — transforms natural-language requests into structured proposals with policy checks and human oversight.
-
INTENTRA Proposal History — a US$25 AI purchase proposal with policy approval, human review, and recorded budget reservation.
-
INTENTRA Intent Contracts — human-approved spending limits, merchant restrictions, and controlled financial authorization.
-
INTENTRA Payments — policy checks, budget reservation, human approval, and a verified US$25 PayPal Sandbox capture.
-
INTENTRA Payment Lifecycle — verified US$25 PayPal Sandbox capture with buyer approval, transaction details, and audit history.
-
INTENTRA Analytics — track payment outcomes, AI proposals, spending policies, and contract budgets in one financial dashboard.
-
INTENTRA tracks verified PayPal Sandbox captures, payment trends, and transaction history, including a successful US$25 test payment.
-
INTENTRA provides account security visibility, session tracking, trusted merchant controls, and human-governed spending authorization.
-
INTENTRA tracks contract approvals, payment events, PayPal verification, and reconciliation through a searchable audit trail.
-
Monitor API and database availability, PayPal Sandbox configuration, trusted merchants, and AI provider settings.
-
Secure account registration and sign-in provide session-based access to INTENTRA's AI purchasing and financial controls.
Inspiration
AI is becoming increasingly capable of understanding what people want and taking action on their behalf. But purchasing introduces a critical question: How can an AI assistant help someone buy something without being given unrestricted control over their money?
Traditional purchasing systems often involve disconnected requests, approvals, budgets, payment processes, and financial records. Meanwhile, AI agents can interpret natural language but should not independently decide when money is authorized or transferred.
We built INTENTRA to bridge that gap.
Our guiding principle is simple:
Intelligent Spending. Human Control.
What it does
INTENTRA is an AI-powered purchasing and financial automation platform that transforms natural-language purchasing requests into structured, policy-controlled workflows.
Instead of allowing AI to make uncontrolled financial decisions, INTENTRA separates intelligent assistance from financial authority.
The purchasing journey follows nine stages:
- Natural-language request: A user describes what they want to purchase.
- AI proposal: The assistant interprets the request and produces structured purchasing information.
- Intent contract: The proposed purchase is associated with a defined purchasing contract.
- Policy evaluation: Deterministic rules evaluate the request against spending constraints.
- Human authorization: A person explicitly reviews and authorizes the action.
- Budget reservation: The system records the financial commitment before payment execution.
- PayPal Sandbox checkout: An approved workflow can proceed through a simulated PayPal payment.
- Verified payment outcome: The application verifies provider results before recording completion.
- Audit history: The workflow maintains traceable records of important decisions and state changes.
This approach is designed to make AI-assisted purchasing more transparent, accountable, and controllable.
How we built it
INTENTRA combines a modern web application, a Python backend, structured data storage, AI-assisted request interpretation, and PayPal payment integration.
Frontend: Next.js, React, and TypeScript provide the purchasing workspace, AI assistant interface, contract management, approval flows, and financial visibility.
Backend: Python and FastAPI implement application APIs, purchasing workflows, authorization boundaries, payment orchestration, and business rules.
Database: PostgreSQL with SQLAlchemy and Alembic provides persistent records and schema management.
AI integration: AI-assisted interpretation converts purchasing requests into structured proposals. The AI supports understanding and preparation, while deterministic application logic governs financial decisions.
Payments: PayPal Sandbox enables testing of the payment lifecycle without transferring real money.
Development and deployment: Docker supports local development, GitHub hosts the source code, and Vercel provides the hosted web application.
A central architectural decision was to keep AI interpretation separate from policy enforcement, human authorization, and payment execution.
Challenges we ran into
Separating AI decisions from financial authority
The most important challenge was ensuring that an AI-generated proposal could not automatically become an authorized payment.
We addressed this by designing distinct workflow stages with explicit authorization requirements and deterministic policy checks.
Managing payment lifecycle states
Payment approval, checkout, capture, and verification are not interchangeable events.
We needed to represent these stages separately so that the interface would not incorrectly present an initiated or approved payment as completed.
Preserving ownership and auditability
Purchasing and financial records require clear ownership boundaries and reliable histories.
We designed the application around owner-scoped resources, persisted workflow state, and audit records that help explain what happened and why.
Building an understandable interface
A financial workflow can be technically correct but still difficult to understand.
We focused on presenting proposals, contracts, approvals, payment states, and financial records in a workspace that helps users understand the consequences of their actions.
Accomplishments that we're proud of
1. A human-governed AI purchasing workflow
INTENTRA demonstrates how natural-language purchasing assistance can operate alongside explicit financial controls rather than bypassing them.
2. PayPal Sandbox integration
We demonstrated a USD 25.00 wireless-mouse test transaction using PayPal Sandbox.
This was a simulated transaction, not a real-money purchase. The demonstration illustrates the connection between purchasing authorization, provider payment processing, verification, and financial records.
3. A structured purchasing lifecycle
The platform brings AI proposals, intent contracts, policy evaluation, human authorization, budget reservation, payment processing, and auditability into one coherent workflow.
4. A full-stack implementation
We built the application using a React/Next.js frontend, FastAPI backend, PostgreSQL persistence, and PayPal Sandbox integration.
5. An auditable approach to financial automation
Rather than optimizing only for how quickly an AI agent can purchase something, we focused on how safely and clearly the system can establish authorization and explain financial outcomes.
What we learned
Building INTENTRA reinforced several important lessons:
- AI intelligence and financial authority should be separate. A model can interpret a request without being trusted to authorize spending.
- Human approval must be meaningful. Authorization should be explicit and tied to the intended transaction.
- Payment verification matters. A checkout response is not automatically proof of a completed payment.
- Auditability is a product feature. Users need to understand not only what happened, but also the decisions that led to it.
- Good design makes complex controls usable. Financial safety mechanisms are most valuable when users can understand and operate them.
What's next for INTENTRA
Our longer-term vision is to expand INTENTRA into a reliable AI-assisted purchasing workspace for individuals and organizations.
Potential future improvements include:
- More advanced purchasing policies and configurable approval workflows.
- Vendor discovery and comparison.
- Improved budget planning and spending analytics.
- Additional payment and financial-service integrations.
- Richer AI assistance for procurement and purchasing decisions.
- Expanded security, reliability, and production-readiness testing.
These are future directions, not claims that every capability is already implemented.
Why INTENTRA matters
As AI agents become more capable of acting on behalf of users, the question is no longer simply whether AI can initiate a purchase.
The more important question is:
Can an AI-powered system make purchasing easier while preserving human authority, financial controls, and accountability?
INTENTRA is our working demonstration of how that can be achieved.
INTENTRA — Intelligent Spending. Human Control. Live website: https://intentra-web.vercel.app Demo video: https://youtu.be/TQfRxiygGl4 Source code: https://github.com/Cgarg547/Intentra
Built With
- alembic
- docker
- fastapi
- github
- next.js
- paypal
- paypal-sandbox
- postgresql
- python
- react
- rest
- sqlalchemy
- typescript
- vercel