PayCommit AI

An autonomous escrow platform using Gemini AI to verify code diffs against milestone criteria and instantly trigger PayPal payouts.

Comment

Inspiration

In traditional freelance and software contracting marketplaces, milestone payments suffer from manual human bottlenecks. Developers often wait days or weeks for clients to manually inspect pull requests, while clients worry about releasing funds before verifying functional delivery. We wanted to pioneer Agentic Commerce—allowing an autonomous AI agent to evaluate software engineering deliverables and execute real-time financial payouts instantly.

What it does

PayCommit AI is an autonomous milestone escrow and payment platform that bridges code verification with instant settlement:

  1. Secured Escrow: Project funds are locked safely via PayPal at contract initiation.
  2. AI Code Verification: When a developer submits a code diff, PR summary, or commit, an objective AI Verification Engine analyzes the work semantically against the milestone criteria.
  3. Automated Settlement: Upon a PASS verdict, the system automatically triggers the PayPal payment API to release funds to the developer with zero manual intervention.
  4. Audit Trail: Every evaluation decision includes an AI reasoning output and timestamped log for transparent dispute protection.

How we built it

  • Backend Engine: Built with Python and FastAPI to orchestrate verification workflows and payment routes.
  • AI Verification Engine: Powered by the Google AI API, engineered to evaluate technical submissions against milestone criteria and output structured JSON decisions.
  • Payments Integration: Integrated with PayPal APIs for milestone escrow locking and automated transaction execution.
  • Frontend Dashboard: Built as a responsive single-page application using Tailwind CSS and modern HTML5/JavaScript.

Challenges we ran into

  • Ensuring AI Output Reliability: Designing prompts that guarantee clean, structured JSON parsing from the AI model so financial triggers execute reliably without syntax errors.
  • Context Evaluation: Balancing strict code verification with fair evaluation so developers are not penalized for minor formatting differences while still ensuring the core milestone features are delivered.

Accomplishments that we're proud of

  • Successfully connecting autonomous LLM code evaluation directly to automated payment execution.
  • Building a sleek, demo-ready dashboard with live status updates, step-by-step progress tracking, and interactive AI logs.
  • Shipping a working end-to-end prototype for the hackathon from scratch.

What we learned

  • How to build safe Agentic Commerce workflows where an AI model is granted execution authority over payment releases based on semantic criteria.
  • Best practices for structuring AI prompts to handle complex software evaluation tasks.

What's next for PayCommit AI

  • GitHub Webhook Integration: Triggering verification automatically whenever a Pull Request is opened or updated on GitHub.
  • Multi-Milestone Contracts: Supporting complex dependency trees where payouts release in automated sequence.
  • Dispute Resolution Protocol: Implementing human-in-the-loop fallback arbitration if either party challenges an AI evaluation.

Built With

Share this project:

Updates

Submission history